The DOI matches its Crossref record; 26 of the 26 references it cites were found in a registry or an index; 1 of the 15 sources its claims cite were checked for support.
BRAKER3: Fully automated genome annotation using RNA-seq and protein evidence with GeneMark-ETP, AUGUSTUS, and TSEBRA
Open accessPublished version of a preprint
Abstract
Gene prediction has remained an active area of bioinformatics research for a long time. Still, gene prediction in large eukaryotic genomes presents a challenge that must be addressed by new algorithms. The amount and significance of the evidence available from transcriptomes and proteomes vary across genomes, between genes, and even along a single gene. User-friendly and accurate annotation pipelines that can cope with such data heterogeneity are needed. The previously developed annotation pipelines BRAKER1 and BRAKER2 use RNA-seq or protein data, respectively, but not both. A further significant performance improvement integrating all three data types was made by the recently released GeneMark-ETP. We here present the BRAKER3 pipeline that builds on GeneMark-ETP and AUGUSTUS, and further improves accuracy using the TSEBRA combiner. BRAKER3 annotates protein-coding genes in eukaryotic genomes using both short-read RNA-seq and a large protein database, along with statistical models learned iteratively and specifically for the target genome. We benchmarked the new pipeline on genomes of 11 species under an assumed level of relatedness of the target species proteome to available proteomes. BRAKER3 outperforms BRAKER1 and BRAKER2. The average transcript-level F1-score is increased by about 20 percentage points on average, whereas the difference is most pronounced for species with large and complex genomes. BRAKER3 also outperforms other existing tools, MAKER2, Funannotate, and FINDER. The code of BRAKER3 is available on GitHub and as a ready-to-run Docker container for execution with Docker or Singularity. Overall, BRAKER3 is an accurate, easy-to-use tool for eukaryotic genome annotation.
Overview
AI snapshot
Written from the paper’s full text.
TL;DR
BRAKER3 is a fully automated eukaryotic genome annotation pipeline that integrates RNA-seq and protein evidence using GeneMark-ETP, AUGUSTUS, and TSEBRA. It outperforms BRAKER1, BRAKER2, and other existing tools in benchmarks, with the largest gains on large and complex genomes.
Key findings
- On 11 species with order-excluded protein databases, BRAKER3 had the highest average accuracy; the average transcript-level F1-score increased by about 20 percentage points relative to BRAKER1/BRAKER2, with the highest increase in *G. gallus* (gene/transcript F1 improved by 55/48 points compared with TSEBRA combining BRAKER1 and BRAKER2).
- Compared with MAKER2 and Funannotate on eight species using close-relatives-included protein databases, BRAKER3’s F1-scores were higher than Funannotate by 10.2 points at exon level, 25.9 at gene level, and 21.6 at transcript level; Funannotate exceeded MAKER2 by 2.2, 3.8, and 4.4 points at exon, gene, and transcript levels, respectively. FINDER completed only 7 of 11 species and its best result (*D. melanogaster*, approximately 15 gene F1 and 11 transcript F1) was much below other methods.
- Transcript-level sensitivity increased with expression: 23% of low-expression transcripts, 55% of medium-expression transcripts, and 76% of highly expressed transcripts were correctly identified. BRAKER3 predicted acceptor/donor splice sites with F1 >87% averaged over species, and start/stop codons with F1 of 70%/76%.
- BUSCO completeness averaged 96% for well-annotated compact genomes, 83.6% for well-annotated large genomes, 91% for other genomes with reference annotation, and 93.7% for three recently sequenced genomes. Runtime ranged from 5 h 37 min for *A. thaliana* to 64 h 16 min for *M. musculus*.
Methods
- BRAKER3 requires a genome sequence, short-read RNA-seq data (BAM, FASTQ, or SRA library IDs), and a protein database. It runs GeneMark-ETP for iterative training and prediction with RNA-seq/protein hints, trains AUGUSTUS on high-confidence genes, and uses TSEBRA to combine GeneMark-ETP and AUGUSTUS predictions while ensuring inclusion of high-confidence genes.
- Benchmarks used 11 species plus three recently sequenced genomes without established reference annotations. Protein databases were prepared as species excluded, order excluded, and close relatives included; comparisons included BRAKER1, BRAKER2, TSEBRA, GeneMark-ETP, MAKER2, FINDER, and Funannotate.
- Accuracy was assessed against reference annotations at exon, gene, and transcript levels using sensitivity, precision, and F1-score; BUSCO was used for genomes lacking reference annotations. Runtime was measured on an HPC node with 48 threads.
Limitations
- The paper states that BRAKER3 cannot be used without RNA-seq evidence, performs worse for weakly expressed transcripts, and currently predicts only protein-coding genes. It also notes benchmark sampling bias toward well-annotated, early-sequenced species, limited RNA-seq input to at most six libraries, eliminated same-order homology evidence, and cautions that reference annotations may fall short and BUSCO does not fully evaluate gene-structure errors or false-positive genes.
- The paper states that the close-relatives-included protein database was manually selected and would not scale well to very large numbers of annotated species.
- Evident from the text: FINDER completed only 7 of 11 species; Funannotate or MAKER2 failed on some large genomes; MAKER2 was trained on the reference annotation, so its reported accuracy is an upper limit; and BRAKER3 had the longest runtime among the compared methods.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
1 of 15 cited sources checked · 0 supported · 0 flagged · accuracy not yet measured
- Findingnone given
The difference in performance is most pronounced for species with large and complex genomes.
“The average transcript-level F1-score is increased by about 20 percentage points on average, whereas the difference is most pronounced for species with large and complex genomes.”
Quote in the paperNot a substantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
The average transcript-level F1-score is increased by about 20 percentage points on average for BRAKER3 compared with BRAKER1 and BRAKER2.
“The average transcript-level F1-score is increased by about 20 percentage points on average, whereas the difference is most pronounced for species with large and complex genomes.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
BRAKER3 outperforms BRAKER1 and BRAKER2.
“BRAKER3 outperforms BRAKER1 and BRAKER2.”
- Also stated in: § BRAKER3 performance assessment · boosted
“Notably, there was a significant improvement of BRAKER3 in comparison with BRAKER1 and BRAKER2 in species with GC-heterogeneous or large genomes (Fig. 3).”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Also stated in: § BRAKER3 performance assessment · boosted
- Findingnone given
BRAKER3 outperforms other existing tools, MAKER2, Funannotate, and FINDER.
“BRAKER3 also outperforms other existing tools, MAKER2, Funannotate, and FINDER.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
When given a large protein database as input (2,588,444 proteins), MAKER2 took 16 h and BRAKER3 took 2.5 h.
“When given a large protein database as input (2,588,444 proteins), MAKER2 took 16 h and BRAKER3 took 2.5 h.”
Quote in the paperSubstantial claim
- Findingown data
The highest exon-level F1-score of ∼88% is achieved for internal CDS.
“The highest exon-level F1-score of ∼88% is achieved for internal CDS (Supplemental Fig. S3).”
Quote in the paperNot a substantial claim - Findingreasoning
The AUGUSTUS prediction using extrinsic evidence of BRAKER3 had a higher sensitivity than GeneMark-ETP at the gene, transcript, and exon level at the cost of lower precisions.
“As a result, this AUGUSTUS prediction using extrinsic evidence of BRAKER3 had a higher sensitivity than GeneMark-ETP at the gene, transcript, and exon level at the cost of lower precisions.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingreasoning
The combined BRAKER3 transcripts have a higher F1-score than either of the combined GeneMark-ETP and AUGUSTUS transcript sets.
“In fact, the combined BRAKER3 transcripts have a higher F1-score than either of the combined GeneMark-ETP and AUGUSTUS transcript sets.”
Quote in the paperSubstantial claim - Findingown data
There was a significant improvement of BRAKER3 in comparison with BRAKER1 and BRAKER2 in species with GC-heterogeneous or large genomes.
“Notably, there was a significant improvement of BRAKER3 in comparison with BRAKER1 and BRAKER2 in species with GC-heterogeneous or large genomes (Fig. 3).”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
The pipelines in order of increased performance are BRAKER1, BRAKER2, TSEBRA combining BRAKER1 and BRAKER2, GeneMark-ETP, and BRAKER3.
“The pipelines in order of increased performance are BRAKER1, BRAKER2, TSEBRA combining BRAKER1 and BRAKER2, GeneMark-ETP, and BRAKER3.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
The pipelines in order of runtime when using the order excluded databases were BRAKER1, GeneMark-ETP, BRAKER2, and BRAKER3.
“The pipelines in order of runtime when using the order excluded databases were BRAKER1, GeneMark-ETP, BRAKER2, and BRAKER3.”
Quote in the paperNot a substantial claim - Findingown data
When run with the species excluded database, BRAKER3 has on average an F1-score that is higher by 0.40, 0.23, and 0.34 at the exon, gene, and transcript levels, respectively, than when BRAKER3 is run with the close relatives included database.
“When run with the species excluded database (Supplemental Tables S6, S7), BRAKER3 has on average an F1-score that is higher by 0.40, 0.23, and 0.34 at the exon, gene, and transcript levels, respectively, than when BRAKER3 is run with the close relatives included database (Supplemental Table S8).”
Quote in the paperSubstantial claim - Findingreasoning
The original TSEBRA tended to overestimate single-exon transcripts in some cases.
“Therefore, the original TSEBRA tended to overestimate single-exon transcripts in some cases.”
Quote in the paperSubstantial claim - Findingown data
In spliced genes, BRAKER3 has more difficulties to predict the initial CDS that contains the start codon than the terminal CDS that contains the stop codon.
“In spliced genes, BRAKER3 has more difficulties to predict the initial CDS that contains the start codon than the terminal CDS that contains the stop codon.”
Quote in the paperNot a substantial claim - Findingown data
In all species, transcripts are much more likely to be correctly identified by BRAKER3 if they are supported by more RNA-seq reads.
“In all species, transcripts are much more likely to be correctly identified by BRAKER3 if they are supported by more RNA-seq reads.”
Quote in the paperNot a substantial claim - Findingnone given
Funannotate failed to run on most of the large OrthoDB-based protein databases.
“Funannotate failed to run on most of the large OrthoDB-based protein databases, and MAKER2 was designed to be used with a smaller protein database, too.”
Quote in the paperSubstantial claim- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
The set of transcripts found by BRAKER3 and GeneMark-ETP, respectively, have large overlaps.
“The set of transcripts found by BRAKER3 and GeneMark-ETP, respectively, have large overlaps (Supplemental Fig. S6).”
Quote in the paperNot a substantial claim - Findingreasoning
By integrating TSEBRA into BRAKER3 and combining sets of gene predictions made by GeneMark-ETP and AUGUSTUS, the final BRAKER3 predictions achieved higher sensitivity and precision than either GeneMark-ETP and AUGUSTUS at both the gene and transcript level.
“By integrating TSEBRA into BRAKER3 and combining sets of gene predictions made by GeneMark-ETP and AUGUSTUS, the final BRAKER3 predictions achieved higher sensitivity and precision than either GeneMark-ETP and AUGUSTUS at both the gene and transcript level.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
The BRAKER3 annotation has higher precision on the exon level than the annotation of GeneMark-ETP and the AUGUSTUS annotation that BRAKER3 produces for all the 11 species.
“Further, the BRAKER3 annotation has higher precision on the exon level than the annotation of GeneMark-ETP and the AUGUSTUS annotation that BRAKER3 produces for all the 11 species (Supplemental Tables S4, S10).”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
The highest performance increase was achieved in G. gallus, in which the BRAKER3 F1-score on gene/transcript level was improved by 55/48 points compared with the combined prediction of BRAKER1 and BRAKER2 generated by TSEBRA.
“The highest performance increase was achieved in G. gallus, in which the BRAKER3 F1-score on gene/transcript level was improved by 55/48 points compared with the combined prediction of BRAKER1 and BRAKER2 generated by TSEBRA (Supplemental Table S4).”
Quote in the paperNot a substantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
Transcripts uniquely predicted by BRAKER3 and not by GeneMark-ETP uncover more of the remaining reference annotation transcripts than vice versa.
“Transcripts uniquely predicted by BRAKER3 and not by GeneMark-ETP uncover more of the remaining reference annotation transcripts than vice versa.”
Quote in the paperSubstantial claim - Findingown data
BRAKER3 is more likely to make an error when predicting an unspliced coding region (a gene with a single CDS feature) than when predicting a CDS of a multi-CDS gene.
“BRAKER3 is more likely to make an error when predicting an unspliced coding region, that is, a gene that has a single coding sequence (CDS) feature, than when predicting a CDS of a multi-CDS gene (Supplemental Fig. S3).”
Quote in the paperNot a substantial claim - Findingown data
When BRAKER3 used the species excluded protein database, the performance measures of the methods using the protein data increased overall.
“When we used the species excluded protein database, which may include very closely related species, the performance measures of the methods using the protein data increased overall (Supplemental Table S6).”
Quote in the paperNot a substantial claim - Findingown data
TSEBRA tends to eliminate false transcripts from either input annotation.
“TSEBRA tends to eliminate false transcripts from either input annotation (for one such example, see Supplemental Figure S2).”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Findingown data
BRAKER3 required only 23% more time on the large order excluded protein databases (averaged over eight species) than on the close relatives included protein databases.
“Nevertheless, BRAKER3 required only 23% more time on these large protein databases (averaged over eight species) (cf. Supplemental Tables S12 and S14).”
Quote in the paperSubstantial claim - Findingown data
When given the relatively small protein database (116,493 proteins), MAKER2 took 2.1 h and BRAKER3 took 3.5 h.
“When given the relatively small protein database as input (116,493 proteins), MAKER2 took 2.1 h and BRAKER3 took 3.5 h.”
Quote in the paperNot a substantial claim - Findingown data
In the medium expression tercile, on average 8.2% of the transcripts are identified by BRAKER3 (and AUGUSTUS) but not by GeneMark-ETP.
“In this expression tercile, on average 8.2% of the transcripts are identified by BRAKER3 (and AUGUSTUS) but not by GeneMark-ETP.”
Quote in the paperSubstantial claim - Findingown data
In C. elegans, about half of the single-exon genes predicted correctly by BRAKER3 are predicted by AUGUSTUS alone.
“In the particular case of C. elegans, even about half of the single-exon genes predicted correctly by BRAKER3 are predicted by AUGUSTUS alone (Supplemental Fig. S6).”
Quote in the paperNot a substantial claim - Findingown data
The BRAKER3 runtime ranges from 5 h 37 min for A. thaliana to 64 h 16 min for M. musculus.
“The BRAKER3 runtime ranges from 5 h 37 min for A. thaliana to 64 h 16 min for M. musculus.”
Quote in the paperNot a substantial claim - Findingown data
Funannotate required on average roughly the same time as BRAKER3.
“Funannotate required on average roughly the same time as BRAKER3 (see Supplemental Table 14).”
Quote in the paperNot a substantial claim - Findingnone given
BRAKER3 predicts acceptor and donor splice sites equally well with an F1-score of >87% averaged over all species.
“BRAKER3 predicts acceptor and donor splice sites equally well with an F1-score of >87% averaged over all species.”
Quote in the paperNot a substantial claim - Findingnone given
The most substantial improvements are observed in species with large and complex genomes.
“The most substantial improvements are observed in species with large and complex genomes.”
Quote in the paperNot a substantial claim - Findingown data
BRAKER3 consistently outperforms Funannotate and MAKER2 at the exon, gene, and transcript levels.
“BRAKER3 consistently outperforms Funannotate and MAKER2 at the exon, gene, and transcript levels (Fig. 5).”
Quote in the paperSubstantial claim - Findingown data
The FINDER annotation pipeline's best performance, an approximately 15 gene F1-score and an approximately 11 transcript F1-score for D. melanogaster, was much below the respective values of the other methods.
“However, its best performance, an approximately 15 gene F1-score and an approximately 11 transcript F1-score for D. melanogaster, was much below the respective values of the other methods (Supplemental Table S9).”
Quote in the paperNot a substantial claim - Findingown data
For nine out of the 11 species, the transcript-level F1-score for predicting spliced transcripts is larger than the respective score for unspliced transcripts.
“It is even the case that for nine out of the 11 species, the transcript-level F1-score for predicting spliced transcripts is larger than the respective score for unspliced transcripts (Supplemental Fig. S4).”
Quote in the paperNot a substantial claim - Findingown data
Of the medium-expressed mouse transcripts, BRAKER3 correctly identifies ∼16% more than GeneMark-ETP.
“Of the medium-expressed mouse transcripts, BRAKER3 correctly identifies ∼16% more than GeneMark-ETP.”
Quote in the paperSubstantial claim - Findingown data
TSEBRA selects most of the single-exon genes predicted by GeneMark-ETP to be in the final set of BRAKER3 genes.
“TSEBRA selects most of the single-exon genes predicted by GeneMark-ETP to be in the final set of BRAKER3 genes (Supplemental Fig. S6).”
Quote in the paperNot a substantial claim - Findingown data
The BRAKER3 transcript sensitivity benefits the most from the integration of AUGUSTUS for medium-expressed transcripts.
“The BRAKER3 transcript sensitivity benefits the most from the integration of AUGUSTUS for medium-expressed transcripts.”
Quote in the paperNot a substantial claim - Findingown data
There was a trade-off in exon sensitivity and precision between BRAKER3 and GeneMark-ETP, with an average difference of approximately eight percentage points in both measures.
“Thus, there was a trade-off in exon sensitivity and precision between the two methods, with an average difference of approximately eight percentage points in both measures (Supplemental Table S4).”
Quote in the paperNot a substantial claim - Findingown data
A linear regression of 19 BRAKER3 whole-genome runtimes yielded the estimate runtime[h] =1.8+2.1⋅genome_size[100MB]; +3.1⋅big_protein_db_used.
“A linear regression of 19 BRAKER3 whole-genome runtimes yielded the estimate runtime[h] =1.8+2.1⋅genome_size[100MB]; +3.1⋅big_protein_db_used.”
Quote in the paperNot a substantial claim - Findingown data
Funannotate exceeded MAKER2 by 2.2, 3.8, and 4.4 points with regards to the F1 measure at exon, gene, and transcript levels, respectively.
“In turn, Funannotate exceeded MAKER2 by 2.2, 3.8, and 4.4 points with regards to the F1 measure at exon, gene, and transcript levels, respectively.”
Quote in the paperNot a substantial claim - Interpretationreasoning
The use of both extrinsic evidence sources together has a clear potential for more accurate gene structure prediction.
“The use of both extrinsic evidence sources together has a clear potential for more accurate gene structure prediction.”
Quote in the paperSubstantial claim - Premisecitation
More precise gene predictions are made when predictions based on statistical models are corrected by extrinsic evidence.
“Therefore, more precise gene predictions are made when predictions based on statistical models are corrected by extrinsic evidence (Stanke et al. 2008; Lomsadze et al. 2014; Brůna et al. 2020).”
Quote in the paperSubstantial claim0 of 3 sources checked- cites Stanke et al. 2008 source
Checked before — its earlier verdict is held back until a new check runs.
- cites Lomsadze et al. 2014 source
Checked before — its earlier verdict is held back until a new check runs.
- cites Brůna et al. 2020
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationreasoning
In contrast to prioritizing sensitivity or achieving high BUSCO completeness scores, BRAKER3 focuses on the precision of its gene predictions.
“In contrast to prioritizing sensitivity or achieving high BUSCO completeness scores, BRAKER3 focuses on the precision of its gene predictions.”
Quote in the paperSubstantial claim - Interpretationreasoning
By itself BUSCO cannot be used to fully evaluate the accuracy and integrity of gene structures.
“Therefore, by itself BUSCO cannot be used to fully evaluate the accuracy and integrity of gene structures.”
Quote in the paperSubstantial claim - Premisecitation
MAKER2 does not train the ab initio models automatically.
“Although MAKER2 can provide training sets for SNAP and AUGUSTUS, it does not train the ab initio models automatically.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Holt and Yandell 2011 source
Checked before — its earlier verdict is held back until a new check runs.
- Study design and conductreasoning
To address the issue of overrepresented annotated species closely related to benchmark species, the authors eliminated homology evidence from the same taxonomic order.
“To address the issue of overrepresented annotated species closely related to our benchmark species, we eliminated homology evidence from the same taxonomic order.”
Quote in the paperSubstantial claim - Interpretationreasoning
Such conclusions from Stanke et al. (2019) do not necessarily carry over to BRAKER3 because the assembly of the transcriptome used in BRAKER3 could benefit from more homogeneous data.
“However, such conclusions do not necessarily carry over to BRAKER3. This is because the assembly of the transcriptome used in BRAKER3 could benefit from more homogeneous data.”
Quote in the paperNot a substantial claim - Interpretationreasoning
The quality of the trained parameters of the gene finders used for MAKER2 can be considered rather as upper limits of what can be expected on new genomes.
“However, the quality of the trained parameters of the gene finders we used for MAKER2 can be considered rather as upper limits of what can be expected on new genomes.”
Quote in the paperNot a substantial claim - Study design and conductreasoning
When run by BRAKER3 on genomic sequences >300 Mbp, TSEBRA removes single-exon genes that are predicted purely ab initio.
“When run by BRAKER3 on genomic sequences >300 Mbp, TSEBRA removes such single-exon genes that are predicted purely ab initio.”
Quote in the paperNot a substantial claim - Interpretationown data
The run-time of BRAKER3 is much less dependent on the protein database size than MAKER2's run-time.
“The run-time of BRAKER3 is much less dependent on the protein database size (here, its decrease was owing to a variable-duration hyperparameter optimization step during training).”
Quote in the paperSubstantial claim - Study design and conductreasoning
Despite the availability of numerous RNA-seq libraries, the authors limited the input for BRAKER3 to a maximum of six RNA-seq libraries.
“Furthermore, despite the availability of numerous RNA-seq libraries (e.g., more than 80,000 for D. melanogaster), we limited the input for BRAKER3 to a maximum of six RNA-seq libraries.”
Quote in the paperSubstantial claim - Interpretationreasoning
GeneMark-ETP enabled the generation of a highly precise set of high-confidence genes to train the AUGUSTUS model.
“GeneMark-ETP enabled the generation of a highly precise set of high-confidence genes to train the AUGUSTUS model.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
- Study design and conductreasoning
In this study, the authors selected well-annotated genomes as benchmarks to predict gene structures, inadvertently introducing a sampling bias, arguably toward species with relatively many related sequenced genomes.
“In our study, we selected well-annotated genomes as benchmarks to predict gene structures, inadvertently introducing a sampling bias, arguably toward species with relatively many related sequenced genomes.”
Quote in the paperSubstantial claim - Study design and conductreasoning
The authors corrected for this bias by restricting the amount and usefulness of available evidence.
“We corrected for this bias by restricting the amount and usefulness of available evidence.”
Quote in the paperSubstantial claim - Interpretationown data
When BRAKER3 uses the larger protein database, it delivers slightly better results.
“Thus, when BRAKER3 uses the larger protein database, it delivers slightly better results.”
Quote in the paperSubstantial claim - Interpretationreasoning
BUSCO does not provide a comprehensive comparison with a mature reference annotation.
“Although BUSCO is commonly used as a standard for assessing the completeness of a proteome, we argue that it does not provide a comprehensive comparison with a mature reference annotation.”
Quote in the paperSubstantial claim - Interpretationown data
Using the large protein database adds an estimated 3.1 h to the runtime.
“Consequently, using the large protein database adds an estimated 3.1 h to the runtime.”
Quote in the paperNot a substantial claim - Premisecitation
Critical to GeneMark-ETP's improvement is a novel approach to generate a training set that has a precision from genes predicted in assembled transcripts and supported by protein evidence.
“Critical to its improvement is a novel approach to generate a training set that has a precision from genes predicted in assembled transcripts and supported by protein evidence.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Brůna et al. 2024
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Probably not a substantial claim. 15
- It may state two separate claims as one. 31
- The quote only partly states this claim. 1
Probably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationFINDER uses RNA-seq assemblies with predicted open reading frames, in conjunction with BRAKER1 and homologous protein predictions.
“FINDER follows an approach similar to BRAKER3. It uses RNA-seq assemblies with predicted open reading frames, in conjunction with BRAKER1 and homologous protein predictions.”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites Banerjee et al. 2021 source
The Implementation section says FINDER assembles RNA-Seq reads into transcripts, predicts their coding regions with GeneMarkS-T, and incorporates protein-derived gene models. It identifies the accompanying gene predictor as BRAKER2, not BRAKER1. (LLM, conf=0.92)
Experimental — accuracy not yet measured
It may state two separate claims as one.
Findingown dataIn each species, BRAKER3 was more precise in predicting exons compared with GeneMark-ETP, which in turn predicted exons more sensitively compared with BRAKER3.
“In each species, BRAKER3 was more precise in predicting exons compared with GeneMark-ETP, which in turn predicted exons more sensitively compared with BRAKER3 (Supplemental Table S4).”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataBRAKER3 shows better performance than Funannotate and MAKER2 for all species and individual metrics, except at the exon level for C. elegans, for which BRAKER3 had a sensitivity 3.3 percentage points lower than Funannotate.
“BRAKER3 shows better performance than Funannotate and MAKER2 for all species and individual metrics, except at the exon level for C. elegans, for which BRAKER3 had a sensitivity 3.3 percentage points lower than the one of Funannotate (Supplemental Table S8).”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataOn average, the BRAKER3 transcript-level sensitivity was improved by approximately three percentage points, and the precision was improved slightly (less than one percentage point) when using the species excluded protein database.
“On average, the BRAKER3 transcript-level sensitivity was improved by approximately three percentage points, and the precision was improved slightly (less than one percentage point).”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataThe relative ranking of the methods and the comparison of BRAKER3 with other methods remain unchanged when using the species excluded protein database.
“However, the relative ranking of the methods and the comparison of BRAKER3 with other methods remain unchanged.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn the group of well-annotated large genomes, AUGUSTUS is 6.6 percentage points more BUSCO-complete than BRAKER3, but BRAKER3 is 30.5 percentage points more precise than AUGUSTUS.
“For example, in the group of well-annotated large genomes, AUGUSTUS is 6.6 percentage points more BUSCO-complete than BRAKER3, but BRAKER3 is 30.5 percentage points more precise than AUGUSTUS.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataAs a trend, a relatively large advantage of BRAKER3 in gene precision stands against a relatively small advantage of AUGUSTUS in BUSCO completeness.
“As a trend, a relatively large advantage of BRAKER3 in gene precision stands against a relatively small advantage of AUGUSTUS in BUSCO completeness.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataThe pattern that BRAKER3-unique transcripts uncover more reference transcripts than GeneMark-ETP-unique transcripts is consistent across all 11 species and applies to both spliced and unspliced transcripts.
“This pattern is consistent across all 11 species and applies to both spliced and unspliced transcripts.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataTSEBRA adds single-exon genes predicted by AUGUSTUS, which increases the percentage of single-exon genes correctly identified by 5.5 percentage points on average.
“However, it adds single-exon genes predicted by AUGUSTUS, which increases the percentage of single-exon genes correctly identified by 5.5 percentage points on average.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataFor each of the groups of species, AUGUSTUS had a higher BUSCO completeness score compared with GeneMark-ETP, which in turn had a higher score compared with BRAKER3.
“For each of the groups of species, AUGUSTUS had a higher BUSCO completeness score compared with GeneMark-ETP, which in turn had a higher score compared with BRAKER3.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataThe average BUSCO completeness (single copy or duplicate) of BRAKER3 is 96% for the group of well-annotated and compact genomes, 83.6% for the group of well-annotated and large genomes, 91% for the group of other genomes with a reference annotation, and 93.7% for the group of three recent genomes.
“The average BUSCO completeness (single copy or duplicate) of BRAKER3 is 96% for the group of well-annotated and compact genomes, 83.6% for the group of well-annotated and large genomes, 91% for the group of other genomes with a reference annotation (M. truncatula, P. tepidariorum, S. lycopersicum), and 93.7% for the group of three recent genomes (Supplemental Fig. S9).”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataOn average, BRAKER3's F1-scores were higher than Funannotate's by 10.2 points at the exon level, 25.9 points at the gene level, and 21.6 points at the transcript level.
“On average, BRAKER3's F1-scores were higher than the ones of Funannotate by 10.2 points at the exon level, 25.9 points at the gene level, and 21.6 points at the transcript level.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataBRAKER3 had the highest sensitivities and precisions for each species at the transcript and gene level but often had a somewhat lower exon-level F1-score compared with GeneMark-ETP.
“BRAKER3 had the highest sensitivities and precisions for each species at the transcript and gene level but often had a somewhat lower exon-level F1-score compared with GeneMark-ETP (Supplemental Table S5).”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Findingown dataThere is a tendency that a tool that predicts more genes genome-wide also reports more of the universal single-copy genes that are counted by BUSCO.
“However, this is also the order of tools sorted by decreasing numbers of predicted genes; that is, there is a (perhaps unsurprising) tendency that a tool that predicts more genes genome-wide also reports more of the universal single-copy genes that are counted by BUSCO.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataWhen averaging over all species, only 23% of low-expression transcripts are correctly identified, 55% of medium-expression transcripts, and 76% of highly expressed transcripts.
“When averaging over all species, only 23% of low-expression transcripts are correctly identified, 55% of medium-expression transcripts, and 76% of highly expressed transcripts.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataThe averaged F1-score for stop and start codons are 76% and 70%, respectively.
“The averaged F1-score for stop and start codons are 76% and 70%, respectively (Supplemental Fig. S5).”
Quote in the paperNot a substantial claimThe quote only partly states this claim.
Findingown dataIn computational experiments with genomes of 11 species, BRAKER3 consistently outperformed the other gene finding tools.
“We have shown that BRAKER3 consistently outperformed the other gene finding tools.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Findingnone givenBRAKER3's performance drops for transcripts that are weakly represented in the RNA-seq data.
“BRAKER3's performance drops for transcripts that are weakly represented in the RNA-seq data.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataAUGUSTUS' average gene- and transcript-level F1-scores of 59.6 and 51.3, respectively, exceeded the F1-scores of the AUGUSTUS predictions in BRAKER1 and BRAKER2 and are slightly lower than the F1-scores of GeneMark-ETP.
“AUGUSTUS’ average gene- and transcript-level F1-scores of 59.6 and 51.3, respectively, exceeded the F1-scores of the AUGUSTUS predictions in BRAKER1 and BRAKER2 and are slightly lower than the F1-scores of GeneMark-ETP (see Supplemental Table S10).”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
It may state two separate claims as one.
PremisecitationGeneMark-ETP integrates both sources of extrinsic evidence in a new workflow that outperforms all previously mentioned methods, particularly in species with large and complex genomes.
“A more integrated approach is the GeneMark-ETP pipeline (Brůna et al. 2024), which integrates both sources of extrinsic evidence in a new workflow that outperforms all previously mentioned methods, particularly in species with large and complex genomes.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Brůna et al. 2024
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
It may state two separate claims as one.
PremisereasoningTSEBRA selects transcripts from BRAKER1 and BRAKER2 annotations, considering the joint extrinsic evidence, and generates a prediction based on both RNA-seq and protein evidence, thus improving F1-scores.
“Therefore, we developed a combiner tool, TSEBRA (Gabriel et al. 2021). It selects transcripts from BRAKER1 and BRAKER2 annotations, considering thereby the joint extrinsic evidence, and therefore generates a prediction based on both RNA-seq and protein evidence, thus improving the F1-scores.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Gabriel et al. 2021 source
Checked before — its earlier verdict is held back until a new check runs.
It may state two separate claims as one.
InterpretationreasoningBRAKER3 annotates protein-coding genes in eukaryotic genomes using both short-read RNA-seq and a large protein database, along with statistical models learned iteratively and specifically for the target genome.
“BRAKER3 annotates protein-coding genes in eukaryotic genomes using both short-read RNA-seq and a large protein database, along with statistical models learned iteratively and specifically for the target genome.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Interpretationown dataBRAKER3 outperforms by a large margin its predecessors BRAKER1 and BRAKER2, as well as publicly available pipelines such as MAKER2, FINDER, and Funannotate.
“BRAKER3 outperforms by a large margin its predecessors BRAKER1 and BRAKER2, as well as publicly available pipelines such as MAKER2, FINDER, and Funannotate.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationPrevious research indicates that gene finders generally show a decreased performance on unspliced transcripts.
“These findings agree with previous research indicating that gene finders generally show a decreased performance on unspliced transcripts (Scalzitti et al. 2020).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Scalzitti et al. 2020 source
Checked before — its earlier verdict is held back until a new check runs.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Premisenone givenFunannotate was initially designed for fungal genomes and has since been further developed to support annotation of larger genomes.
“The Funannotate pipeline, which was not described in an article, was initially designed as a pipeline for analyzing fungal genomes; however, it has since been further developed to support the annotation of larger genomes as well.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites https://github.com/nextgenusfs/funannotate
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationMAKER2 was designed to be used with a smaller protein database.
“Funannotate failed to run on most of the large OrthoDB-based protein databases, and MAKER2 was designed to be used with a smaller protein database, too.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Holt and Yandell 2011 source
Checked before — its earlier verdict is held back until a new check runs.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
PremisecitationIn Stanke et al. (2019), a manual selection of libraries from different experiments and tissues was compared with an automatic sampling from diverse libraries for complementary data; the latter yields, on average, slightly better performances in BRAKER1 when the number of reads was similar.
“In their study, a manual selection of libraries from different experiments and tissues was compared with an automatic sampling from diverse libraries for complementary data. The latter yields, on average, slightly better performances in BRAKER1 when the number of reads was similar.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Stanke et al. (2019) source
Checked before — its earlier verdict is held back until a new check runs.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
Interpretationnone givenCurrently, BRAKER3 predicts only protein-coding genes.
“Currently, BRAKER3 predicts only protein-coding genes.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
PremisereasoningTSEBRA heavily relies on intron position information, which can make it challenging to evaluate single-exon transcripts.
“TSEBRA heavily relies on intron position information, which can make it challenging to evaluate single-exon transcripts.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningBRAKER3 greatly benefited from the high accuracy of GeneMark-ETP and managed to exceed the sensitivity and precision on gene and transcript levels even further.
“Here, BRAKER3 greatly benefited from the high accuracy of GeneMark-ETP and managed to exceed the sensitivity and precision on gene and transcript levels even further.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
- The authors state this as an assumption: The target species' proteome has an assumed level of relatedness to available proteomes.
- Rests on an assumption: The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
- Rests on an assumption: The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
It may state two separate claims as one.
InterpretationreasoningFor these gene sets, a better BUSCO score comes at the expense of more false-positive genes, and BRAKER3 weights the trade-off between fewer missing genes and fewer false-positive genes in favor of the latter.
“In conclusion, for these gene sets, a better BUSCO score comes at the expense of more false-positive genes, and BRAKER3 weights the trade-off between fewer missing genes and fewer false-positive genes in favor of the latter.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
PremisecitationGeneMark-ETP benefited from GC-content-specific model training and from estimating species-specific repeat penalties.
“The method also benefited from the GC-content-specific model training and from estimating species-specific repeat penalties.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Brůna et al. 2024
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
It may state two separate claims as one.
Interpretationnone givenBUSCO's design does not encompass the detection of various gene structure errors, nor does it impose penalties for the presence of false-positive genes.
“Specifically, BUSCO's design does not encompass the detection of various gene structure errors, nor does it impose penalties for the presence of false-positive genes.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductnone givenWith the diverse species selection, it was not feasible to use RNA-seq sets that are comparable across different species.
“With our diverse species selection, it was not feasible to use RNA-seq sets that are comparable across different species.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisereasoningThe self-training of GeneMark.hmm models has to be performed outside of MAKER2.
“Also, the self-training of GeneMark.hmm models (Lomsadze et al. 2005) has to be performed outside of MAKER2.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Lomsadze et al. 2005
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Probably not a substantial claim. It may state two separate claims as one.
Study design and conductnone givenThe long-established approach of comparing predicted annotations with reference annotations allows measurement of grave as well as subtle errors and is informative in absolute terms if the reference annotations are of high quality.
“This long-established approach to benchmark gene prediction methods allows measurement of grave as well as subtle errors and is informative in absolute terms if the reference annotations are of high quality.”
Quote in the paperSubstantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
It may state two separate claims as one.
InterpretationreasoningBRAKER3 combines sets of transcripts and can be seen as an ensemble learning approach that improves each of its inputs.
“BRAKER3 combines these sets of transcripts and can be seen as an ensemble learning approach that improves each of its inputs.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Interpretationnone givenThe species for the close relatives included database were manually selected, and the procedure would not scale well when very large numbers of species are annotated.
“It should be noted that the species for the close relatives included database were manually selected, and the procedure would not scale well when very large numbers of species are annotated.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
InterpretationreasoningSome reference annotations may fall short of the desired accuracy, potentially leading to an underestimation of accuracy or a bias in the analysis.
“It should be noted that some of the reference annotations may fall short of the desired accuracy, potentially leading to an underestimation of accuracy or a bias in our analysis.”
Quote in the paperNot a substantial claim- The authors state this as an assumption: The reference annotations used as ground truth are sufficiently accurate for benchmarking.
Probably not a substantial claim. The paper may give no reason why it holds.
Interpretationnone givenBRAKER3 cannot be used without RNA-seq evidence.
“Notably, BRAKER3 cannot be used without RNA-seq evidence.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationStanke et al. (2019) investigated the impact of an RNA-seq sampling method on the performance of BRAKER1.
“However, Stanke et al. (2019) investigated the impact of an RNA-seq sampling method on the performance of BRAKER1.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Stanke et al. (2019) source
Checked before — its earlier verdict is held back until a new check runs.
It may state two separate claims as one.
Study design and conductreasoningThe figures for MAKER2 do not include the considerable times for training gene finders and/or transcriptome assembly, whereas BRAKER3 performs these steps as part of the pipeline.
“When comparing these runtimes, one has to consider that the figures for MAKER2 do not include the considerable times for training gene finders and or for transcriptome assembly. In contrast, BRAKER3 performs these steps as part of the pipeline.”
Quote in the paperSubstantial claim
The authors state this as an assumption.
The reference annotations used as ground truth are sufficiently accurate for benchmarking.
“This long-established approach to benchmark gene prediction methods allows measurement of grave as well as subtle errors and is informative in absolute terms if the reference annotations are of high quality.”
The authors state this as an assumption.
The target species' proteome has an assumed level of relatedness to available proteomes.
“We benchmarked the new pipeline on genomes of 11 species under an assumed level of relatedness of the target species proteome to available proteomes.”
The authors state this as an assumption.
The order excluded protein database is a conservative assumption.
“we performed experiments with the particularly favorable case of the species excluded database and with the rather conservative assumption of the order excluded database.”
The paper relies on this without support.
The five or six randomly selected short-read RNA-seq libraries per species are sufficient and representative for annotation.
“For each species, we retrieved genome assemblies, five or six randomly selected short-read RNA-seq libraries from NCBI's Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra) (for detailed list, see Supplemental Table S15), a protein database, and a reference genome annotation (for detailed list, see Supplemental Table S1).”
The paper relies on this without support.
The OrthoDB-based protein databases from the broad clade contain useful homologous informants for the target species.
“The protein database is a FASTA file with proteins from the broad clade of the target genome in question, for example, a subset from the partitioning of OrthoDB that we provide (see Supplemental Methods).”
The paper relies on this without support.
Modern genome annotation pipelines depend on an increasing number of tools, which can make their installation and maintenance difficult.
“One problem with modern genome annotation pipelines is their dependence on an increasing number of tools, which can make their installation and maintenance difficult.”
The paper relies on this without support.
BUSCO uses a set of genes that are well conserved and, therefore, arguably relatively easy to predict.
“Moreover, BUSCO uses a set of genes that are well conserved and, therefore, arguably relatively easy to predict.”
The paper relies on this without support.
A comprehensive gold-standard annotation for a set of test genomes representing the whole space of eukaryotic genomes does not exist.
“However, a comprehensive gold-standard annotation for such a set of genomes does not exist.”
The paper relies on this without support.
The error rates in eukaryotic gene annotation remain significant.
“As long as the error rates in eukaryotic gene annotation remain significant, a trade-off between the numbers of false-positive and false-negative genes has to be made.”
Citations
Cites
Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype
Kim, D., Paggi, J. M., Park, C., Bennett, C., & Salzberg, S. L. (2019) · Nature Biotechnology · doi:10.1038/s41587-019-0201-4 · Cited by 15,022
BUSCO: assessing genome assembly and annotation completeness with single-copy orthologs
Simão, F. A., Waterhouse, R. M., Ioannidis, P., Kriventseva, E. V., & Zdobnov, E. M. (2015) · Bioinformatics · doi:10.1093/bioinformatics/btv351 · Cited by 13,344
Near-optimal probabilistic RNA-seq quantification
Bray, N. L., Pimentel, H., Melsted, P., & Pachter, L. (2016) · Nature Biotechnology · doi:10.1038/nbt.3519 · Cited by 9,723
RepeatModeler2 for automated genomic discovery of transposable element families
Flynn, J. M., Hubley, R., Goubert, C., Rosen, J., Clark, A. G., Feschotte, C., & Smit, A. F. (2020) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1921046117 · Cited by 3,848
Korf, I. F. (2004) · BMC Bioinformatics · doi:10.1186/1471-2105-5-59 · Cited by 3,160
Leinonen, R., Sugawara, H., & Shumway, M. F. (2010) · Nucleic Acids Research · doi:10.1093/nar/gkq1019 · Cited by 2,852
AUGUSTUS: ab initio prediction of alternative transcripts
Stanke, M., Keller, O., Gunduz, I., Hayes, A., Waack, S., & Morgenstern, B. (2006) · Nucleic Acids Research · doi:10.1093/nar/gkl200 · Cited by 2,829
Self-citationSingularity: Scientific containers for mobility of compute
Kurtzer, G. M., Sochat, V., & Bauer, M. W. (2017) · PLOS ONE · doi:10.1371/journal.pone.0177459 · Cited by 2,740
Using native and syntenically mapped cDNA alignments to improve de novo gene finding
Stanke, M., Diekhans, M., Baertsch, R., & Haussler, D. (2008) · Bioinformatics · doi:10.1093/bioinformatics/btn013 · Cited by 2,605
Self-citationHolt, C., & Yandell, M. (2011) · BMC Bioinformatics · doi:10.1186/1471-2105-12-491 · Cited by 2,118
Transcriptome assembly from long-read RNA-seq alignments with StringTie2
Kovaka, S., Zimin, A. V., Pertea, G. M., Razaghi, R., Salzberg, S. L., & Pertea, M. (2019) · Genome biology · doi:10.1186/s13059-019-1910-1 · Cited by 2,000
Brůna, T., Hoff, K. J., Lomsadze, A., Stanke, M., & Borodovsky, M. Y. (2021) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqaa108 · Cited by 1,643
Self-citationBRAKER1: Unsupervised RNA-Seq-Based Genome Annotation with GeneMark-ET and AUGUSTUS
Hoff, K. J., Lange, S., Lomsadze, A., Borodovsky, M. Y., & Stanke, M. (2015) · Bioinformatics · doi:10.1093/bioinformatics/btv661 · Cited by 1,154
Self-citationHershberg, R., & Petrov, D. A. (2008) · Annual Review of Genetics · doi:10.1146/annurev.genet.42.110807.091442 · Cited by 1,056
Gene identification in novel eukaryotic genomes by self-training algorithm
Lomsadze, A. (2005) · Nucleic Acids Research · doi:10.1093/nar/gki937 · Cited by 985
Self-citationIntegration of mapped RNA-Seq reads into automatic training of eukaryotic gene finding algorithm
Lomsadze, A., Burns, P. D., & Borodovsky, M. (2014) · Nucleic Acids Research · doi:10.1093/nar/gku557 · Cited by 612
Self-citationGeneMark-EP+: eukaryotic gene prediction with self-training in the space of genes and proteins
Brůna, T., Lomsadze, A., & Borodovsky, M. Y. (2020) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqaa026 · Cited by 542
Self-citationOrthoDB v11: annotation of orthologs in the widest sampling of organismal diversity
Kuznetsov, D. A., Tegenfeldt, F., Manni, M., Seppey, M., Berkeley, M., Kriventseva, E. V., & Zdobnov, E. M. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkac998 · Cited by 530
Identification of protein coding regions in RNA transcripts
Tang, S., Lomsadze, A., & Borodovsky, M. Y. (2015) · Nucleic Acids Research · doi:10.1093/nar/gkv227 · Cited by 506
Self-citationTSEBRA: transcript selector for BRAKER
Gabriel, L., Hoff, K. J., Brůna, T., Borodovsky, M., & Stanke, M. (2021) · BMC Bioinformatics · doi:10.1186/s12859-021-04482-0 · Cited by 410
Self-citationLilue, J., Doran, A. G., Fiddes, I. T., Abrudan, M., Armstrong, J., Bennett, R., Chow, W., Collins, J., Collins, S., Czechanski, A., et al. (2018) · Nature Genetics · doi:10.1038/s41588-018-0223-8 · Cited by 267
Self-citationGeneMark-ETP significantly improves the accuracy of automatic annotation of large eukaryotic genomesPreprint
Brůna, T., Lomsadze, A., & Borodovsky, M. Y. (2024) · Genome Research · doi:10.1101/gr.278373.123 · Cited by 172
Self-citationGalba: genome annotation with miniprot and AUGUSTUS
Brůna, T., Li, H., Guhlin, J. G., Honsel, D., Herbold, S., Stanke, M., Nenasheva, N., Ebel, M., Gabriel, L., & Hoff, K. J. (2023) · BMC Bioinformatics · doi:10.1186/s12859-023-05449-z · Cited by 105
Self-citationA benchmark study of ab initio gene prediction methods in diverse eukaryotic organisms
Scalzitti, N., Jeannin-Girardon, A., Collet, P., Poch, O., & Thompson, J. D. (2020) · BMC Genomics · doi:10.1186/s12864-020-6707-9 · Cited by 96
Banerjee, S., Bhandary, P., Woodhouse, M., Sen, T. Z., Wise, R. P., & Andorf, C. M. (2021) · BMC Bioinformatics · doi:10.1186/s12859-021-04120-9 · Cited by 35
VARUS: sampling complementary RNA reads from the sequence read archive
Stanke, M., Bruhn, W., Becker, F., & Hoff, K. J. (2019) · BMC Bioinformatics · doi:10.1186/s12859-019-3182-x · Cited by 18
Self-citation
Flagged first, then by citations
Cited by
Sorted by citations
GeneMark-ETP significantly improves the accuracy of automatic annotation of large eukaryotic genomesPreprint
Brůna, T., Lomsadze, A., & Borodovsky, M. Y. (2024) · Genome Research · doi:10.1101/gr.278373.123 · Cited by 172
Self-citationIntegrated biotechnological and AI innovations for crop improvement
Li, G., An, L., Yang, W., Yang, L., Wei, T., Shi, J., Wang, J., Doonan, J. H., Xie, K., Fernie, A. R., et al. (2025) · Nature · doi:10.1038/s41586-025-09122-8 · Cited by 150
Cartney, A. M. M., Formenti, G., Mouton, A., Panis, D. N. D., Marins, L. S., Leitão, H. G., Diedericks, G., Kirangwa, J., Morselli, M., Salces-Ortiz, J., et al. (2024) · npj Biodiversity · doi:10.1038/s44185-024-00054-6 · Cited by 80
Brůna, T., Lomsadze, A., & Borodovsky, M. Y. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.01.13.524024 · Cited by 66
Self-citationTiberius: end-to-end deep learning with an HMM for gene prediction
Gabriel, L., Becker, F., Hoff, K. J., & Stanke, M. (2024) · Bioinformatics · doi:10.1093/bioinformatics/btae685 · Cited by 51
Self-citationLewin, T. D., Liao, I. J.-Y., & Luo, Y. (2024) · Molecular Biology and Evolution · doi:10.1093/molbev/msae172 · Cited by 49
A telomere-to-telomere genome assembly of the cultivated peanut
Wang, X., Sun, Z., Qi, F., Zhou, Z., Du, P., Shi, L., Dong, W., Huang, B., Suoyi, H., Pavan, S., et al. (2024) · Molecular Plant · doi:10.1016/j.molp.2024.12.001 · Cited by 38
An episodic burst of massive genomic rearrangements and the origin of non-marine annelids
Vargas-Chávez, C., Benítez-Álvarez, L., Martínez-Redondo, G. I., Álvarez-González, L., Salces-Ortiz, J., Eleftheriadi, K., Escudero, N., Guiglielmoni, N., Flot, J.-F., Novo, M., et al. (2025) · Nature Ecology & Evolution · doi:10.1038/s41559-025-02728-1 · Cited by 35
Adaptation repeatedly uses complex structural genomic variation
Gompert, Z., Feder, J. L., Parchman, T. L., Planidin, N. P., Whiting, F. J. H., & Nosil, P. (2025) · Science · doi:10.1126/science.adp3745 · Cited by 35
A multispecies foundation model for genome function prediction, gene annotation and sequence designPreprint
Boshar, S., Evans, B., Tang, Z., Adel, Y., Bornachot, Y., Sidbon, S., Lorbeer, F. K., Picard, A., Rajesh, C., Seitz, E., et al. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2025.12.22.695963 · Cited by 35
Highly accurate ab initio gene annotation with ANNEVO
Zhang, P., Xu, T., Wang, S., Yang, X., Sun, P., Jia, P., Lin, J., Wang, B., Zhang, Y., Meng, D., et al. (2026) · Nature Methods · doi:10.1038/s41592-026-03036-7 · Cited by 34
Genome analyses and breeding of polyploid crops
Lin, C. T., Bao, Z., Kong, Q., Lassois, L., Stein, N., Huang, S., & Zhou, Q. (2025) · Nature Plants · doi:10.1038/s41477-025-02088-5 · Cited by 33
The ivory lncRNA regulates seasonal color patterns in buckeye butterflies
Fandino, R. A., Brady, N. K., Chatterjee, M., McDonald, J. M. C., Livraghi, L., van der Burg, K. R. L., Mazo‐Vargas, A., Markenscoff-Papadimitriou, E., & Reed, R. D. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2403426121 · Cited by 31
Cao, Y., Feng, X., Ding, B., Huo, H., Abdullah, M., Hong, J., Jiang, L., Wang, H., Li, R., Cai, Y., et al. (2024) · Plant Communications · doi:10.1016/j.xplc.2024.101238 · Cited by 30
Spatial variation in the mutation rate within the plant shoot apical meristem
Amundson, K., Marimuthu, M. P. A., Nguyen, O., Sarika, K., DeMarco, I. J., Phan, A., Henry, I. M., & Comai, L. (2025) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2514507122 · Cited by 30
PSAURON: a tool for assessing protein annotation across a broad range of species
Sommer, M. J., Zimin, A. V., & Salzberg, S. L. (2025) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqae189 · Cited by 29
Quest for Orthologs in the Era of Biodiversity Genomics
Langschied, F., Bordin, N., Cosentino, S., Fuentes-Palacios, D., Glover, N., Hiller, M., Hu, Y., Huerta-Cepas, J., Coelho, L. P., Iwasaki, W., et al. (2024) · Genome Biology and Evolution · doi:10.1093/gbe/evae224 · Cited by 26
Sato, Y., Bex, R., van den Berg, G. C. M., Santhanam, P., Höfte, M., Seidl, M. F., & Thomma, B. P. H. J. (2025) · Nature Communications · doi:10.1038/s41467-025-61986-6 · Cited by 26
Navigating Eukaryotic Genome Annotation Pipelines: A Route Map to Using BRAKER, Galba, and TSEBRA
Brůna, T., Gabriel, L., & Hoff, K. J. (2025) · Methods in molecular biology · doi:10.1007/978-1-0716-4583-3_4 · Cited by 26
Self-citationYano, R., Li, F., Hiraga, S., Takeshima, R., Kobayashi, M., Toda, K., Umehara, Y., Kajiya‐Kanegae, H., Iwata, H., Kaga, A., et al. (2025) · Nature Genetics · doi:10.1038/s41588-025-02113-5 · Cited by 24
Francis, B. A., Gozashti, L., Costello, K. R., Kasahara, T., Harringmeyer, O. S., Lilue, J., Wu, T., Zoltowska, K. M., Helmy, M., Kato, T., et al. (2025) · Nature Genetics · doi:10.1038/s41588-025-02367-z · Cited by 24
Chromosome-level genome assembly of the sacoglossan sea slug Elysia timida (Risso, 1818)
Männer, L., Schell, T., Spies, J., Galià‐Camps, C., Baranski, D., Hamadou, A. B., Gerheim, C., Neveling, K., Helfrich, E. J. N., & Greve, C. (2024) · BMC Genomics · doi:10.1186/s12864-024-10829-7 · Cited by 24
Fusion, fission, and scrambling of the bilaterian genome in BryozoaPreprint
Lewin, T. D., Liao, I. J.-Y., Chen, M.-E., Bishop, J., Holland, P. W. H., & Luo, Y. (2025) · Genome Research · doi:10.1101/gr.279636.124 · Cited by 23
Schell, T., Greve, C., & Podsiadłowski, L. (2025) · Frontiers in Zoology · doi:10.1186/s12983-025-00561-7 · Cited by 22
Global Genomics of Man-O'-War (Physalia) Reveals Ocean Surface Biodiversity
Church, S. H., Abedon, R. B., Ahuja, N., Anthony, C. J., Destanović, D., Ramirez, D. A., Rojas, L. M., Albinsson, M. E., Trasobares, I. Á., Bergemann, R., et al. (2024) · Current Biology · doi:10.1016/j.cub.2025.05.066 · Cited by 22
Phoronid genome supports a monophyletic Lophophorata
Lewin, T. D., Sakagami, T., Kao, L.-J., Liao, I. J.-Y., Chiu, Y., Piñón-González, V. M., Li, J.-Y., Tin, K.-X., Lu, M. J., Endo, K., et al. (2025) · Current Biology · doi:10.1016/j.cub.2025.10.020 · Cited by 22
Cai, Y., Gao, X., Mao, J., Liu, Y., Tong, L., Chen, X., Liu, Y., Kou, W., Chang, C., Foster, T., et al. (2024) · Nature Communications · doi:10.1038/s41467-024-54428-2 · Cited by 22
Graph-based pangenomes and pan-phenome provide a cornerstone for eggplant biology and breeding
Gaccione, L., Toppino, L., Bolger, M., Schmidt, M., Tassone, M. R., Sulli, M. M., Idahl, E., Alonso, D., Aprea, G., Ferrante, P., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-64866-1 · Cited by 21
Zhang, F., Wang, Y., Lin, Y., Wang, H., Wu, Y., Ren, W., Wang, L., Yang, Y., Zheng, P., Wang, S., et al. (2024) · Molecular Horticulture · doi:10.1186/s43897-024-00083-6 · Cited by 21
Telomere-to-telomere Genome Assembly of the Clubroot Pathogen Plasmodiophora Brassicae
Javed, M. A., Mukhopadhyay, S., Normandeau, É., Brochu, A., & Pérez‐López, E. (2024) · Genome Biology and Evolution · doi:10.1093/gbe/evae122 · Cited by 21
Annotation matters: the effect of structural gene annotation on orthology inference
Prieto-Baños, S., Nevers, Y., Altenhoff, A., Warwick Vesztrocy, A., Dessimoz, C., & Glover, N. M. (2025) · Bioinformatics · doi:10.1093/bioinformatics/btaf365 · Cited by 19
Jackson, D. J., Cerveau, N., & Posnien, N. (2024) · Frontiers in Zoology · doi:10.1186/s12983-024-00538-y · Cited by 19
A pangenome reveals LTR repeat dynamics as a major driver of genome evolution in Chenopodium
Jaggi, K. E., Krak, K., Štorchová, H., Mandák, B., Marcheschi, A. K., Belyayev, A., Jellen, E. N., Sproul, J. S., Jarvis, D. E., & Maughan, P. J. (2025) · The Plant Genome · doi:10.1002/tpg2.70010 · Cited by 19
Guiglielmoni, N., Villegas, L. I., Kirangwa, J., & Schiffer, P. H. (2024) · Frontiers in Genetics · doi:10.3389/fgene.2024.1308527 · Cited by 18
Phytopathogen Effector Biology in the Burgeoning AI Era
Jones, D., & Raffaele, S. (2025) · Annual Review of Phytopathology · doi:10.1146/annurev-phyto-121823-081033 · Cited by 17
Toward an integrated omics approach for plant biosynthetic pathway discovery in the age of AI
Reinhardt, J. K., Craft, D., & Weng, J. (2025) · Trends in Biochemical Sciences · doi:10.1016/j.tibs.2025.01.010 · Cited by 17
Rissi, D. V., Ijaz, M., & Baschien, C. (2024) · Journal of Fungi · doi:10.3390/jof10090632 · Cited by 16
Conservation of bilaterian genome structure is the exception, not the rule
Lewin, T. D., Liao, I. J.-Y., & Luo, Y. (2025) · Genome biology · doi:10.1186/s13059-025-03732-1 · Cited by 16
An X-linked sex determination mechanism in cannabis and hop
Carey, S. B., Bentz, P. C., Lovell, J. T., Akozbek, L. M., Myers, Z., Korani, W., Havill, J. S., Padgitt‐Cobb, L. K., Lynch, R. C., Allsing, N. A., et al. (2026) · Nature Communications · doi:10.1038/s41467-026-73233-7 · Cited by 16
AMIR: a multi-omics data platform for Asteraceae plants genetics and breeding research
Liu, D., Luo, C., Dai, R., Huang, X., Chen, X., He, L., Mao, H., Li, J., Zhang, L., Yang, Q., et al. (2024) · Nucleic Acids Research · doi:10.1093/nar/gkae833 · Cited by 15
McIntyre, J., Morrison, A. A., Maitland, K., Berger, D., Price, D. R. G., Dougan, S., Grigoriadis, D., Tracey, A., Holroyd, N. E., Bull, K., et al. (2025) · PLoS Pathogens · doi:10.1371/journal.ppat.1012820 · Cited by 15
Liu, H., Zhao, H., Zhang, Y., Li, X., Zuo, Y., Wu, Z., Jin, K., Xian, W., Wang, W., Ning, W., et al. (2024) · Journal of Integrative Plant Biology · doi:10.1111/jipb.13765 · Cited by 14
Kulkarni, S., Klementz, B. C., & Sharma, P. P. (2024) · Journal of Heredity · doi:10.1093/jhered/esae074 · Cited by 13
The genome assembly of the duckweed fern, Azolla caroliniana
Song, M. J., Rizzieri, Y. C., Li, F., Freund, F. D., Escalona, M., Toffelmier, E. M., Miller, C., Shaffer, H. B., Nguyen, O., Marimuthu, M. P. A., et al. (2025) · Journal of Heredity · doi:10.1093/jhered/esaf022 · Cited by 13
dsRNAmax: a multi-target chimeric dsRNA designer for safe and effective crop protection
Fletcher, S. J., Lawrence, J., Sawyer, A., Manzie, N., Gardiner, D. M., Mitter, N., & Brosnan, C. A. (2025) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqaf064 · Cited by 13
Thier, O. D., Lebel, M., M.Tawfeeq, M., Faure, R., Dru, P., Blanchoud, S., Alié, A., Brown, F. D., Flot, J., & Tiozzo, S. (2025) · GigaScience · doi:10.1093/gigascience/giaf097 · Cited by 13
Building better genome annotations across the tree of lifePreprint
Freedman, A. H., & Sackton, T. B. (2025) · Genome Research · doi:10.1101/gr.280377.124 · Cited by 13
Rebalancing the seed proteome following deletion of vicilin-related genes in pea ( Pisum sativum L.)
Rayner, T., Saalbach, G., Vickers, M. J., Paajanen, P., Martins, C., Wouters, R. H. M., Chinoy, C. N., Mulholland, F., Bal, M., Isaac, P. G., et al. (2024) · Journal of Experimental Botany · doi:10.1093/jxb/erae518 · Cited by 12
Liu, S., Shi, C., Chen, C., Tan, Y., Tian, Y., Macqueen, D. J., & Li, Q. (2025) · Cell Reports · doi:10.1016/j.celrep.2025.115697 · Cited by 12
Brůna, T., Sreedasyam, A., Harder, A. M., & Lovell, J. T. (2026) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqag011 · Cited by 12
Self-citationTransposable elements hitchhike on Starships across fungal genomes
Griem-Krey, H., Sant’Ana, J. D. F., Oggenfuss, U., Calegari-Alves, Y. P., MARQUES, A. L., Berger, M., Santi, L., Beys‐da‐Silva, W. O., & Habig, M. (2026) · Nature Communications · doi:10.1038/s41467-026-69410-3 · Cited by 12
Alves, R. M., de Abreu, V. A. C., de Oliveira, R. P., Almeida, J. V. D. A., Oliveira, M. D. M., Silva, S. R., Paschoal, A. R., de Almeida, S. S., de Souza, P. A. F., Ferro, J. A., et al. (2024) · GigaScience · doi:10.1093/gigascience/giae027 · Cited by 12
Plant-associated fungi co-opt ancient antimicrobials for host manipulation
Mesny, F., Wolf, V., López-Moral, A., Kraege, A., Punt, W., Liu, S., Zhu, J., Park, J., Sato, Y., & Thomma, B. P. H. J. (2026) · Science Advances · doi:10.1126/sciadv.aec1406 · Cited by 11
Oriowo, T. O., Chrysostomakis, I., Martín, S., Kukowka, S., Brown, T., Winkler, S., Myers, E. W., Böhne, A., & Stange, M. (2024) · GigaScience · doi:10.1093/gigascience/giae116 · Cited by 11
Cell type-specific immune regulation under symbiosis in a facultatively symbiotic coral
Valadez-Ingersoll, M., Rivera, H. E., Da‐Anoy, J., Kanke, M., Gómez-Campo, K., Martinez-Rugerio, M. I., Metz, S., Sweet, M. J., Kwan, J., Hekman, R. M., et al. (2025) · The ISME Journal · doi:10.1093/ismejo/wraf132 · Cited by 11
Constraints on chromosome evolution revealed by the 229 chromosome pairs of the Atlas blue butterfly
Wright, C. J., Absolon, D. E., Gascoigne‐Pees, M., Vila, R., Lawniczak, M., & Blaxter, M. (2025) · Current Biology · doi:10.1016/j.cub.2025.08.032 · Cited by 11
Karbstein, K., Choudhary, N., Xie, T., Tomasello, S., Wagner, N. D., Barke, B. H., Paetzold, C., Bradican, J. P., Preick, M., Himmelbach, A., et al. (2025) · The Plant Journal · doi:10.1111/tpj.70390 · Cited by 11
Cookbook for plant genome sequences
de Oliveira, J. A. V. S., Choudhary, N., Meckoni, S. N., Nowak, M. S., Hagedorn, M., & Pucker, B. (2026) · BMC Genomics · doi:10.1186/s12864-026-12623-z · Cited by 11
Kataoka, K., Sanno, R., Gaczorek, T. S., Bhattarai, U. R., Ito, Y., Inoue, S., Yura, K., Asahi, T., Ylla, G., Mito, T., et al. (2026) · G3 Genes Genomes Genetics · doi:10.1093/g3journal/jkag036 · Cited by 11
Chromosome-scale assembly of the wild cereal relative Elymus sibiricus
Shen, W., Liu, B., Guo, J., Yang, Y., Li, X., Chen, J., & Dou, Q. (2024) · Scientific Data · doi:10.1038/s41597-024-03622-4 · Cited by 11
Chromosome-level genome assemblies of sunflower oilseed and confectionery cultivars
Yi, L., Bao, H., Wu, Y., Mu, Y., Du, C., Peng, J., Yan, X., Chen, Y., & Yu, H. (2025) · Scientific Data · doi:10.1038/s41597-024-04097-z · Cited by 11
A telomere-to-telomere genome assembly of Camellia nitidissima
Wang, X.-F., Liu, T.-J., Feng, T., Huang, H.-R., Zou, P., Xiao, W., Wu, X., Chai, S., & Yan, H. (2025) · Scientific Data · doi:10.1038/s41597-025-05157-8 · Cited by 11
Chromosome-scale and haplotype-resolved genome assembly of Populus trichocarpa
Gao, W., Wang, S., Jiang, T., Hu, H. K., Gao, R., Zhou, M., & Wang, G. (2025) · Horticulture Research · doi:10.1093/hr/uhaf012 · Cited by 10
Gene novelty and gene family expansion in the early evolution of Lepidoptera
Hoile, A. E., Holland, P. W. H., & Mulhair, P. O. (2025) · BMC Genomics · doi:10.1186/s12864-025-11338-x · Cited by 10
InsectBase 3.0: a comprehensive multi-omics resource for insects
Wang, Z., Chen, H., Jin, S., Li, Z., Yang, M., Xiao, Y., Zhao, C., Zhou, T., Li, F., Liu, Y., et al. (2025) · Nucleic Acids Research · doi:10.1093/nar/gkaf1248 · Cited by 10
Genomic evidence of a complex supergene system linking dispersal to social polymorphism
Mona, S., Gay, E. J., Taupenot, A., Ducancel, J., Laso‐Jadart, R., Helleu, Q., Chifflet‐Belle, P., Téodori, E., Aury, J., Xiong, Z., et al. (2025) · Current Biology · doi:10.1016/j.cub.2025.10.065 · Cited by 10
Wang, Z., Zhou, J., Pan, J., Cheng, W., Fang, J., Lv, Q., Lin, X., Cheng, W., Zhang, L., & Cheng, K. (2023) · Horticulture Research · doi:10.1093/hr/uhad274 · Cited by 10
Haplotype-resolved genome assembly of the tetraploid potato cultivar Désirée
Godec, T., Beier, S., Rodriguez‐Granados, N. Y., Sasidharan, R., Abdelhakim, L. O. A., Teige, M., Usadel, B., Gruden, K., & Petek, M. (2025) · Scientific Data · doi:10.1038/s41597-025-05372-3 · Cited by 10
Σιαπέρας, Ρ., Taxeidis, G., Gioti, Α., Nikolaivits, E., & Topakas, E. (2025) · Environmental Pollution · doi:10.1016/j.envpol.2025.126386 · Cited by 9
Recuerda, M., Kraemer, S., Rosoni, J. R. R., Repenning, M., Browne, M., Cataudela, J. F., Giacomo, A. S. D., Kopuchian, C., & Campagna, L. (2025) · Evolution · doi:10.1093/evolut/qpaf188 · Cited by 9
Arms races between selfish genetic elements and their host defence in termites
Qiu, B., Elsner, D., & Korb, J. (2026) · Nature Communications · doi:10.1038/s41467-026-69550-6 · Cited by 9
Annotating genomes at increased scale and resolution
Ji, H. J., Pertea, M., & Salzberg, S. L. (2026) · Nature Reviews Genetics · doi:10.1038/s41576-026-00937-3 · Cited by 9
Yin, H. C., Jiang, D. H., Yu, T. F., Jiang, X. J., & Liu, D. (2024) · Frontiers in Cellular and Infection Microbiology · doi:10.3389/fcimb.2024.1432422 · Cited by 9
Yisilam, G., Zheng, E., Li, C., Zhang, Z., Su, Y., Chu, Z., Li, P., & Tian, X. (2025) · Acta Botanica Yunnanica · doi:10.1016/j.pld.2025.01.001 · Cited by 9
Molino, J. V. D., Oliver, A., Sethuram, H., Kang, K., Saucedo, B., Diaz, C. J., Gupta, A., Jen, L. J., Torres-Tiji, Y., Hidasi, N., et al. (2025) · Algal Research · doi:10.1016/j.algal.2025.104034 · Cited by 9
Chromosome-level genome assembly of soybean aphid
Qiu, S., Wu, N., Sun, X., Xue, Y. J., & Xia, J. (2025) · Scientific Data · doi:10.1038/s41597-025-04711-8 · Cited by 9
Lind, A., McDonald, N. A., Gerrick, E. R., Bhatt, A. S., & Pollard, K. S. (2025) · Genome Research · doi:10.1101/gr.279080.124 · Cited by 9
The genomic insights of intertidal adaptation in Bryopsis corticulans
Xu, P., Liu, X., Lei, K., Li, K., Wang, W., & Jiao, Y. (2025) · New Phytologist · doi:10.1111/nph.70083 · Cited by 8
Lin, Y. E., Wu, C., Wu, Y., & Chaw, S. (2025) · Plants · doi:10.3390/plants14091335 · Cited by 8
Nearly telomere-to-telomere genome assembly of the L. edodes diploid genome
Yoshitake, K., Shirasawa, K., Kojima, K. K., Asakawa, S., Tanaka, N., & Kurokochi, H. (2025) · Fungal Genetics and Biology · doi:10.1016/j.fgb.2025.104005 · Cited by 8
Groh, J. S., Ackerman, G., Wenzel, M., & Coop, G. (2025) · Current Biology · doi:10.1016/j.cub.2025.09.061 · Cited by 8
Rare microbial relict sheds light on an ancient eukaryotic supergroup
Valt, M., Pánek, T., Mirzoyan, S., Tice, A. K., Jones, R. E., Dohnálek, V., Doležal, P., Mikšátko, J., Rotterová, J., Hrubá, P., et al. (2025) · Nature · doi:10.1038/s41586-025-09750-0 · Cited by 8
Chromosomal fusions trigger rediploidization of autopolyploid genomes
Xie, C., Ma, Z., Zhou, C., Ma, K., Wang, H., Wu, J., Zhou, Y., Lu, Y., Ji, D., Gu, X., et al. (2026) · Nature · doi:10.1038/s41586-026-10439-1 · Cited by 8
Ferguson, S., Bar-Ness, Y. D., Borevitz, J., & Jones, A. (2024) · BMC Genomics · doi:10.1186/s12864-024-10810-4 · Cited by 8
Hemara, L. M., Chatterjee, A., Yeh, S., Chen, R., Hilario, E., Lievre, L. L., Crowhurst, R. N., Bohne, D., Arshed, S., Patterson, H. R., et al. (2024) · Plant Cell & Environment · doi:10.1111/pce.15189 · Cited by 8
Jones, B. M., Webb, A. E., Geib, S. M., Sim, S. B., Schweizer, R. M., Branstetter, M. G., Evans, J. D., & Kocher, S. D. (2024) · Molecular Biology and Evolution · doi:10.1093/molbev/msae229 · Cited by 8
Cho, Y., Seo, C. W., Cho, H. J., Jin, Y., Lupala, A. S., Shim, S. H., & Lim, Y. W. (2025) · BMC Genomics · doi:10.1186/s12864-025-11542-9 · Cited by 8
Annotation of protein-coding genes in 49 diatom genomes from the Bacillariophyta clade
Nenasheva, N., Pitzschel, C., Webster, C., Hart, A. J., Wegrzyn, J., Bengtsson, M. M., & Hoff, K. J. (2025) · Scientific Data · doi:10.1038/s41597-025-05306-z · Cited by 8
Self-citationZhao, Q., Yin, Z., & Hou, Z. (2025) · Journal of Animal Science and Biotechnology/Journal of animal science and biotechnology · doi:10.1186/s40104-024-01141-1 · Cited by 7
Faster adaptation but slower divergence of X chromosomes under paternal genome elimination
Baird, R. B., Hitchcock, T. J., Ševčı́k, J., Monteith, K. M., Gardner, A., Ross, L., & Mongue, A. J. (2025) · Nature Communications · doi:10.1038/s41467-025-60114-8 · Cited by 7
White paper: standards for handling and analyzing plant pan-genomes
Heuermann, M. C., Barros, P., Beier, S., Gundlach, H., Álvarez-Jarreta, J., Hassani‐Pak, K., König, P., Fiebig, A., Godec, T., Gruden, K., et al. (2025) · F1000Research · doi:10.12688/f1000research.166538.1 · Cited by 7
Gao, P., Lian, L., Feng, W., Ma, Y., Lin, J., Qin, L., Hao, S., Zhao, H., Liu, X., Yuan, J., et al. (2025) · Genome biology · doi:10.1186/s13059-025-03782-5 · Cited by 7
Pelosi, J. A., Davenport, R., Kuo, L., Gray, L. N., Dant, A. J., Kim, E. H., Li, F., Dlugosch, K. M., Krabbenhoft, T. J., Barbazuk, W. B., et al. (2025) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2504773122 · Cited by 7
Coexisting With Humans: Genomic and Behavioral Consequences in a Small and Isolated Bear Population
Fabbri, G., Biello, R., Gabrielli, M., Vilaça, S. T., Sammarco, B., Fuselli, S., Santos, P., Ancona, L., Peretto, L., Padovani, G., et al. (2025) · Molecular Biology and Evolution · doi:10.1093/molbev/msaf292 · Cited by 7
Ancient gene linkages and ultraconserved elements disentangle Acari interrelationships
Bhoi, J. P., Mule, R., Savale, N., Sharma, P. P., & Kulkarni, S. (2026) · iScience · doi:10.1016/j.isci.2025.114616 · Cited by 7
Plant genome assembly and annotation
Michael, T. P. (2026) · Current Opinion in Plant Biology · doi:10.1016/j.pbi.2026.102859 · Cited by 7
Rapid mid-Cretaceous diversification of squid and cuttlefish preceded radiation into coastal niches
Sánchez, G., Fernández‐Álvarez, F. Á., Bernal, A., Heath-Heckman, E., Lami, R., McFall‐Ngai, M., Nishiguchi, M. K., Nyholm, S. V., Simakov, O., Allcock, A. L., et al. (2026) · Nature Ecology & Evolution · doi:10.1038/s41559-026-03009-1 · Cited by 7
A geothermal amoeba sets a new upper temperature limit for eukaryotes
Rappaport, H. B., Petek-Seoane, N. A., Tyml, T., Niblo, J. K., Mikus, F., Gilbert, N. E., LaButti, K. M., Ani, G., MacVicar, E., Shepherd, R. M., et al. (2026) · Cell · doi:10.1016/j.cell.2026.08.043 · Cited by 7
Bryant, M. J., Coello, A. M., Glendening, A. M., Hilliman, S. A., Jara, C., Pring, S. S., Rivera, A., Membreño, J. S., Nigro, L. M., Pauloski, N., et al. (2024) · Genome Biology and Evolution · doi:10.1093/gbe/evae097 · Cited by 7
Metadata
- Authors
- Gabriel, L., Brůna, T., Hoff, K. J., Ebel, M., Lomsadze, A., Borodovsky, M., & Stanke, M.
- Affiliations
- Universität Greifswald (DE)
- Lawrence Berkeley National Laboratory (US)
- Georgia Institute of Technology (US)
- Corresponding author
- Hoff, Katharina J.
- Lomsadze, Alexandre
- ORCID
- Venue
- Genome Research
- Publisher
- Cold Spring Harbor Laboratory
- Published
- 2024-05
- Volume
- 34
- Issue
- 5
- Pages
- 769-777
- ISSN
- 1088-9051 (ISSN-L)
- 1549-5469
- PMC
- PMC11216308
- OpenAlex
- W4399560644
- Cited by
- 676
- References
- 26
- Citations by year
- 2026: 360 · 2025: 245 · 2024: 63
- Type
- posted content
- Language
- en
- Open access
- green
- www.ncbi.nlm.nih.gov
- Licence
- cc-by-nc-sa
- Locations
- Genome Research (publishedVersion)
- PubMed Central (publishedVersion, cc-by-nc-sa)
- PubMed (publishedVersion)
- OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) (submittedVersion, other-oa)
- HighWire Press Open Archive (submittedVersion)
- Funders
- U.S. National Institutes of Health (GM128145)
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Chromosomal and Genetic Variations
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › RNA and protein synthesis mechanisms
- Keywords
- BRAKER3, genome annotation, gene prediction, RNA-seq, Augustus, eukaryotic genomes, transcriptome, proteome, annotation pipeline, Funannotate, Finder, Docker containers
- MeSH terms
- RNA-Seq, Algorithms, Animals, Humans, Software, Genome, Computational Biology, Genomics, Molecular Sequence Annotation, Transcriptome
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-11-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
