The DOI matches its Crossref record; 39 of the 39 references it cites were found in a registry or an index; none of the 17 sources its claims cite is checked for support yet.
Modeling of RNA-seq fragment sequence bias reduces systematic errors in transcript abundance estimation
Open accessPublished version of a preprint
Abstract
We find that current computational methods for estimating transcript abundance from RNA-seq data can lead to hundreds of false-positive results. We show that these systematic errors stem largely from a failure to model fragment GC content bias. Sample-specific biases associated with fragment sequence features lead to misidentification of transcript isoforms. We introduce alpine, a method for estimating sample-specific bias-corrected transcript abundance. By incorporating fragment sequence features, alpine greatly increases the accuracy of transcript abundance estimates, enabling a fourfold reduction in the number of false positives for reported changes in expression compared with Cufflinks. Using simulated data, we also show that alpine retains the ability to discover true positives, similar to other approaches. The method is available as an R/Bioconductor package that includes data visualization tools useful for bias discovery.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
30 Own result · 11 Own interpretation · 9 Not checked yet · 4 Own method · 2 No basis given
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The lightweight quantification methods generated the same kind of isoform-switching errors as Cufflinks and RSEM (Supplementary Figures 14-15), and as MISO 27, a statistical method for estimating isoform percentages within multi-isoform genes”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“the drop in coverage for samples from center 1 results in a shift in expression estimates from NM003367 to NM207291, which does not include the high GC exon”
plaindirectSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Overall, the mixing ratio recovery for all methods was higher for poly-A selected samples (65-75%) compared to ribosomal RNA depleted samples (45-65%).”
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The expected mixing ratios in the ABRF dataset are accurate ground truth for assessing measurement accuracy.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Correcting for fragment sequence bias using alpine reduced the number of extreme predicted changes in isoform percent when comparing across center”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“RSEM had similar rates of differentially expressed transcripts across center”
plainn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Using only read start bias terms in the alpine model did not provide visible improvements”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The samples from center 1 have dramatically reduced representation of high GC fragments compared to center 2, even after adjusting for differences due to random hexamer priming bias”
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Strong GC content bias was found for some samples for all public datasets examined.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We found that fragment GC content dependence was strongly associated with the location of library preparation, and not with the location of sequencing”
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The SEQC samples prepared at different sites are comparable except for library preparation location.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“When restricting positives to those transcripts with Benjamini-Hochberg adjusted p value less than 1% and additionally requiring that the log2 fold change be above a threshold (0.5, 1, or 2), alpine consistently had a lower percent of false positives out of the set of 5,676 transcripts than all other methods”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The fragment sequence models accurately capture the drops in coverage that were not captured by the read start sequence models”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“as random permutations of samples resulted in no transcripts with differential expression.”
plainn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Out of 5,716 transcripts from genes with two isoforms in which at least one isoform had FPKM greater than 0.1, 566 transcripts reported differential expression across center according to Cufflinks estimated abundances, at a target FDR of 1%.”
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“For genes with one or more isoforms reported differentially expressed, the regions of the isoform that were exclusive to one or the other isoform had much higher GC-content (mode at 70% compared to 50%) than expected by chance (Figure 2b, Wilcoxon p < 0.0001).”
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We find that current computational methods for estimating transcript abundance from RNA-seq data can lead to hundreds of false-positive results.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We found that including bias terms for fragment GC and GC stretches resulted in a fourfold decrease in the number of false positives at a target FDR of 1%: Cufflinks reported 562 differentially expressed transcripts, while alpine reported 141”
plaindirectn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In a comparison of expression estimates across protocol, alpine reported the fewest number of false positives, controlling at 1% FDR and at 1% FWER”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The ABRF samples prepared with different protocols are comparable except for the protocol.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Using simulated data, we also show that alpine retains the ability to discover true positives, similar to other approaches.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Performing a t-test on log2(FPKM + 1) values from Cufflinks across centers, and filtering on Benjamini-Hochberg adjusted p values for a target 1% false discovery rate (FDR) resulted in 2,510 transcripts out of 25,588 (10%) with average FPKM greater than 0.1 reported as differentially expressed”
plainn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“While the sequence features we included in our model provided substantial improvements over existing methods”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“When comparing across center, 619 out of 6,761 genes with multiple isoforms and average FPKM greater than 0.1 for one or more isoform had changes in the reported major isoform across centers using FPKM values estimated by Cufflinks”
plainn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Note that four of the methods evaluated, Cufflinks, kallisto, Salmon, and Sailfish all were run with read start sequence bias correction turned on (Supplementary Note) and obtained similar results”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Sample-specific biases associated with fragment sequence features lead to mis-identification of transcript isoforms.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
None of the sources cited for this claim have been checked yet.
In the paper
“At the transcript level, sample-specific biases that current methods correct for include the fragment length distribution induced by size selection, positional bias along the transcript due to RNA degradation and mRNA selection techniques, and sequence-based bias in read start positions arising from the differential binding efficiency of random hexamer primers2,12-16”
What the sources say
- Not checked yetcites [2] Li, B., Ruotti, V., Stewart, R.M., Thomson, J.A. & Dewey, C.N. RNA-Seq gene expression estimation with read mapping uncertainty. Bioinformatics 26, 493–500 (2010). source
- Not checked yetcites [12] Li, J., Jiang, H. & Wong, W.H. Modeling non-uniformity in short-read rates in RNA-Seq data. Genome Biol. 11, R50 (2010). source
- Not checked yetcites [13] Hansen, K.D., Brenner, S.E. & Dudoit, S. Biases in Illumina transcriptome sequencing caused by random hexamer priming. Nucleic Acids Res. 38, e131 (2010). source
- Not checked yetcites [14] Roberts, A., Trapnell, C., Donaghey, J., Rinn, J.L. & Pachter, L. Improving RNA-Seq expression estimates by correcting for fragment bias. Genome Biol. 12, R22–R14 (2011). source
- Not checked yetcites [15] Nicolae, M., Mangul, S., Maˇndoiu, I.I. & Zelikovsky, A. Estimation of alternative splicing isoform frequencies from RNA-Seq data. Algorithms Mol. Biol. 6, 9 (2011). source
- Not checked yetcites [16] Li, W. & Jiang, T. Transcriptome assembly and isoform expression level estimation from biased RNA-Seq reads. Bioinformatics 28, 2914–2921 (2012). source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
None of the sources cited for this claim have been checked yet.
In the paper
“However, RNA-seq data can suffer from sample-specific biases as a result of RNA extraction and library preparation steps3–5.”
What the sources say
- Not checked yetcites [3] 't Hoen, P.A. et al. Reproducibility of high-throughput mRNA and small RNA sequencing across laboratories. Nat. Biotechnol. 31, 1015–1022 (2013). source
- Not checked yetcites [4] Su, Z. & SEQC/MAQC-III Consortium A comprehensive assessment of RNA-seq accuracy, reproducibility and information content by the Sequencing Quality Control Consortium. Nat. Biotechnol. 32, 903–914 (2014). source
- Not checked yetcites [5] Li, S. et al. Detecting and correcting systematic variation in large-scale RNA sequencing data. Nat. Biotechnol. 32, 888–895 (2014). source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We note that our findings reflect general systematic errors and not just differences induced by batch effects.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“because fold changes induced by technical bias are often larger than those potentially of biological interest.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, we expect that nearly all of the 2,510 transcripts reporting differential expression are false positives”
hedgedn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The 30 GEUVADIS samples from the two sequencing centers have no true biological differences across centers, so across-center differential expression is technical.
In the paper
“Additionally, stretches of high GC content sequences within a fragment can have an influence on whether a fragment will be amplified and sequenced 23.”
What the sources say
- Not checked yetcites [23] Hron, T., Pajer, P., Pačes, J., Baru˚tneˇk, P. & Elleder, D. Hidden genes in birds. Genome Biol. 16, 164 (2015). source
plainconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We show that these systematic errors stem largely from a failure to model fragment GC content bias.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“There are likely to be many incorrectly reported major isoforms and biased abundance estimates for experiments that show strong dependence of the fragment rate on fragment GC content (e.g. Figure 2e), unless these are explicitly corrected for using methods that model fragment sequence bias.”
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“a comparison of C/D and A/B for the different methods can be used to generate a measure of accuracy in estimating transcript abundance.”
§ Modeling of RNA-seq fragment sequence bias reduces systematic errors in transcript abundance estimation > Online Methods > Accuracy in mixing ratios for ABRF samples
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The C and D samples in the ABRF dataset were made by mixing the A and B samples in 3:1 and 1:3 ratios, respectively.
In the paper
“Indeed, ’t Hoen et al.3 observed that slight differences in average GC content across samples lead to differences in quantification for transcripts.”
What the sources say
- Not checked yetcites ’t Hoen et al.3
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- It may state two separate claims as one. 21
- Probably not a substantial claim. 5
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“alpine had the highest sensitivity for a given specificity in the confounded design (Figure 3g-h), while methods performed similarly for the balanced design”
plainn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The Polyester-simulated RNA-seq data, with fragment GC dependence estimated from GEUVADIS samples, adequately represents real RNA-seq bias for sensitivity analysis.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“While MISO performed better than the existing transcript abundance estimators in consistency of isoform identification, alpine reported less than half of the large isoform switches across center reported by MISO”
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“we confirmed other findings 21, that the GC content effect was often nonlinear and highly sample-specific, therefore requiring sample-specific bias estimation using smooth functions of fragment GC content.”
What the sources say
- Not checked yetcites [21] Benjamini, Y. & Speed, T.P. Summarizing and correcting the GC content bias in high-throughput sequencing. Nucleic Acids Res. 40, e72 (2012). source
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Fitting the fragment sequence model does require more computational effort than other models, though our implementation ran in comparable time to the cuffquant step of the Cufflinks suite”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The model that also included the information about GC stretches was more predictive than the model with just the fragment GC content, although only slightly so.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“alpine had the highest accuracy in estimating the true expression values compared to the other methods (Supplementary Figures 18-19), with low median absolute error in estimating the percent of isoform abundance for genes with two isoforms, comparable to RSEM”
plainn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“When quantifying the number of transcripts whose C/D ratio was within 10% of the expected value, RSEM had the highest recovery for both protocols, though alpine also had consistently high recovery (within 5% of the top method).”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The expected mixing ratios in the ABRF dataset are accurate ground truth for assessing measurement accuracy.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The models that included fragment GC content doubled the predictive power of the read start bias models, both the Cufflinks VLMM read start bias model implemented within alpine and mseq, which performed similarly”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In general, alpine greatly reduced across-center significant differences while within-center coefficient of variation of abundance estimates remained the same as for Cufflinks”
plaindirectn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“We observed little change in the shape of fragment GC dependence across protocol, and a strong effect in the positional bias, with poly-A selected samples having highest coverage at the 3’ end of transcripts, as reported by the ABRF study authors29”
What the sources say
- Not checked yetcites [29] Li, S. et al. Multi-platform assessment of transcriptome profiling using RNA-seq in the ABRF next-generation sequencing study. Nat. Biotechnol. 32, 915–925 (2014). source
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The ABRF samples prepared with different protocols are comparable except for the protocol.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“By incorporating fragment sequence features, alpine greatly increases the accuracy of transcript abundance estimates, enabling a fourfold reduction in the number of false positives for reported changes in expression compared with Cufflinks.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Using a more conservative Bonferroni correction, Cufflinks reported 157 differentially expressed transcripts across center with FWER of 1%, while alpine reported only 37.”
plainn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In contrast, scenarios with total or partial confounding, unknown batches, and sample-specific deviations within batches are typical in RNA-seq experiments and not represented by the balanced design simulation.”
§ Modeling of RNA-seq fragment sequence bias reduces systematic errors in transcript abundance estimation > Online Methods > Simulation
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may state two separate claims as one.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“As the process of discarding pairs of reads would result in unequal final library size, the simulation process was repeated, after first scaling the initial target library size higher such that the final library size for the experiment would be equal to 60 million paired-end reads in expectation.”
§ Modeling of RNA-seq fragment sequence bias reduces systematic errors in transcript abundance estimation > Online Methods > Simulation
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“The alpine software was run with bias correction terms for fragment length, relative position, and fragment GC content, and other methods were run with their bias correction arguments turned on, including a positional bias correction term for RSEM (Supplementary Note).”
§ Modeling of RNA-seq fragment sequence bias reduces systematic errors in transcript abundance estimation > Online Methods > Specificity analysis in ABRF samples
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
In the paper
“For DNA-seq, this bias is best corrected by modeling at the scale of the fragment21.”
What the sources say
- Not checked yetcites [21] Benjamini, Y. & Speed, T.P. Summarizing and correcting the GC content bias in high-throughput sequencing. Nucleic Acids Res. 40, e72 (2012). source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“Note that the GC content curves used in the simulation (Figure 2e) were estimated after removing the read start bias using the Cufflinks VLMM, and so the fragment sequence effect observed is not a proxy for read start bias.”
§ Modeling of RNA-seq fragment sequence bias reduces systematic errors in transcript abundance estimation > Online Methods > Simulation
plainn = 30- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
None of the sources cited for this claim have been checked yet.
In the paper
“Even so, it is common to observe extreme variability in the coverage of RNA-seq fragments along transcripts that is purely technical and sample-specific17 and not explained by current bias models, which confounds current methods designed to identify and quantify transcripts18”
What the sources say
- Not checked yetcites [17] Lahens, N.F. et al. IVT-seq reveals extreme bias in RNA sequencing. Genome Biol. 15, R86 (2014). source
- Not checked yetcites [18] Hayer, K.E., Pizarro, A., Lahens, N.F., Hogenesch, J.B. & Grant, G.R. Benchmark analysis of algorithms for determining and quantifying full-length mRNA splice forms from RNA-seq data. Bioinformatics 31, 3938–3945 (2015). source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The problem we identify holds for within-sample transcript abundance estimates for samples from a single center and for samples in small-scale experiments.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“Here we demonstrated specificity using data prepared by different sequencing centers, and sensitivity using simulation.”
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Note that the balanced design with blocking factor represents the best-case scenario for the competing methods, as the batches were known exactly and the residual degrees of freedom was high, such that the batch effects on transcript abundance estimates could be precisely estimated and removed by the linear model.”
§ Modeling of RNA-seq fragment sequence bias reduces systematic errors in transcript abundance estimation > Online Methods > Simulation
plaindirectn = 15- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“A GC content bias is often observed in high-throughput sequencing data, with fragments of certain GC content under-represented, and can be partially attributed to PCR amplification during library preparation20.”
What the sources say
- Not checked yetcites [20] Aird, D. et al. Analyzing and minimizing PCR amplification bias in Illumina sequencing libraries. Genome Biol. 12, R18 (2011). source
plainconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The large number of transcripts with false positive differences in abundance reported across protocol, in the range of 10-14,000 at 1% FDR for all methods out of 28,000, suggests that none of the existing methods evaluated including alpine can remove the bulk of systematic bias seen across protocol.”
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The ABRF samples prepared with different protocols are comparable except for the protocol.
The paper relies on this without support.
The 30 GEUVADIS samples from the two sequencing centers have no true biological differences across centers, so across-center differential expression is technical.
“We downloaded 30 RNA-seq samples from the GEUVADIS Project of lymphoblastoid cell lines derived from the Toscani population, 15 of which were sequenced at one center and 15 at another19”
The paper relies on this without support.
The Polyester-simulated RNA-seq data, with fragment GC dependence estimated from GEUVADIS samples, adequately represents real RNA-seq bias for sensitivity analysis.
“we performed a sensitivity analysis using an RNA-seq fragment simulator, Polyester 28, to generate 30 samples with fragment GC content dependence estimated from the 30 GEUVADIS samples, after having removed read start sequence bias”
The paper relies on this without support.
The expected mixing ratios in the ABRF dataset are accurate ground truth for assessing measurement accuracy.
“The relationship of the samples in the ABRF dataset allowed an assessment of the measurement accuracy in terms of the methods’ recovery of expected mixing ratios”
The paper relies on this without support.
The SEQC samples prepared at different sites are comparable except for library preparation location.
“including libraries of the same samples that were prepared and sequenced at three different sites and libraries prepared at a separate site and only sequenced at the three sites”
The paper relies on this without support.
The ABRF samples prepared with different protocols are comparable except for the protocol.
“we downloaded a subset of samples from the ABRF Next-Generation Sequencing Study29 that were prepared using either ribosomal RNA depletion or poly-A selection”
The paper relies on this without support.
The IVT transcripts with high unpredictable coverage are representative of transcripts affected by fragment sequence bias.
“We focused our analysis on 64 of the IVT transcripts as exhibiting “high unpredictable coverage”17.”
The paper relies on this without support.
The poly-A selection protocol does not capture all non-coding transcripts.
“as the poly-A selection protocol is not designed to capture all the non-coding transcripts”
The paper relies on this without support.
The C and D samples in the ABRF dataset were made by mixing the A and B samples in 3:1 and 1:3 ratios, respectively.
“As the C and D samples in the ABRF dataset were created by mixing 3:1 and 1:3 ratios of the A and B samples respectively”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“When uncorrected, these biases can lead to, e.g., high false positive rates in differential expression studies [5].”
From Salmon provides fast and bias-aware quantification of transcript expression (2017) · cites this paper as [5]
In the citing paper
“However, existing methods for transcriptome-wide abundance estimation, both alignment-based and alignment-free, lack sample-specific bias models rich enough to capture important effects, like fragment GC content bias.”
From Salmon provides fast and bias-aware quantification of transcript expression (2017) · cites this paper as [5]
In the citing paper
“it has also been observed that the GC-content of the entire fragment can play a substantial role in the likelihood that it will be selected for sequencing [5].”
From Salmon provides fast and bias-aware quantification of transcript expression (2017) · cites this paper as [5]
In the citing paper
“the accuracy of these estimates is hindered by incomplete annotations, as well as by limitations of short-read sequencing, coverage, and genomic biases like GC content (84, 85).”
From Transcriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder (2018) · cites this paper as [85]
Citations
Cites
STAR: ultrafast universal RNA-seq aligner
Dobin, A., Davis, C. A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., Batut, P., Chaisson, M., & Gingeras, T. R. (2013) · Bioinformatics · doi:10.1093/bioinformatics/bts635 · Cited by 51,709
RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome
Li, B., & Dewey, C. N. (2011) · BMC Bioinformatics · doi:10.1186/1471-2105-12-323 · Cited by 22,179
Differential expression analysis for sequence count data
Anders, S., & Huber, W. (2010) · Genome Biology · doi:10.1186/gb-2010-11-10-r106 · Cited by 15,581
Trapnell, C., Williams, B. A., Pertea, G., Mortazavi, A., Kwan, G., van Baren, M. J., Salzberg, S. L., Wold, B. J., & Pachter, L. (2010) · Nature Biotechnology · doi:10.1038/nbt.1621 · Cited by 15,556
Salmon provides fast and bias-aware quantification of transcript expression
Patro, R., Duggal, G., Love, M. I., Irizarry, R. A., & Kingsford, C. (2017) · Nature Methods · doi:10.1038/nmeth.4197 · Cited by 12,700
Self-citationKim, D., Pertea, G., Trapnell, C., Pimentel, H., Kelley, R., & Salzberg, S. L. (2013) · Genome Biology · doi:10.1186/gb-2013-14-4-r36 · Cited by 12,668
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
Software for Computing and Annotating Genomic Ranges
Lawrence, M. F., Huber, W. K., Pagès, H., Aboyoun, P., Carlson, M. R. J., Gentleman, R. C., Morgan, M. T., & Carey, V. J. (2013) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1003118 · Cited by 4,475
Orchestrating high-throughput genomic analysis with Bioconductor
Huber, W. K., Carey, V. J., Gentleman, R. C., Anders, S., Carlson, M. R. J., Carvalho, B. S., Bravo, H. C., Davis, S. R., Gatto, L., Girke, T., et al. (2015) · Nature Methods · doi:10.1038/nmeth.3252 · Cited by 3,614
Self-citationNormalization of RNA-seq data using factor analysis of control genes or samples
Risso, D., Ngai, J., Speed, T. P., & Dudoit, S. (2014) · Nature Biotechnology · doi:10.1038/nbt.2931 · Cited by 2,052
Research works printed but not linked: 1
- Patro, R., Duggal, G. & Kingsford, C. Accurate, fast, and model-aware transcript expression quantification with Salmon. bioRxiv (2015).
Cited by
Sorted by citations
Salmon provides fast and bias-aware quantification of transcript expression
Patro, R., Duggal, G., Love, M. I., Irizarry, R. A., & Kingsford, C. (2017) · Nature Methods · doi:10.1038/nmeth.4197 · Cited by 12,700
Self-citationDifferential analyses for RNA-seq: transcript-level estimates improve gene-level inferences
Soneson, C., Love, M. I., & Robinson, M. D. (2015) · F1000Research · doi:10.12688/f1000research.7563.1 · Cited by 3,929
Self-citationTranscriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder
Gandal, M. J., Zhang, P., Hadjimichael, E., Walker, R. L., Chen, C., Liu, S., Won, H., van Bakel, H., Varghese, M., Wang, Y., et al. (2018) · Science · doi:10.1126/science.aat8127 · Cited by 1,247
Voorwerk, L., Slagter, M., Horlings, H. M., Sikorska, K., de Vijver, K. K. V., de Maaker, M., Nederlof, I., Kluin, R. J. C., Warren, S., Ong, S., et al. (2019) · Nature Medicine · doi:10.1038/s41591-019-0432-4 · Cited by 932
A general and flexible method for signal extraction from single-cell RNA-seq data
Risso, D., Perraudeau, F., Gribkova, S., Dudoit, S., & Vert, J.-P. (2018) · Nature Communications · doi:10.1038/s41467-017-02554-5 · Cited by 718
Metadata
- Authors
- Love, M. I., Hogenesch, J. B., & Irizarry, R. A.
- Affiliations
- Harvard University (US)
- University of Pennsylvania (US)
- Corresponding author
- Irizarry, Rafael A
- Venue
- Nature Biotechnology
- Published
- 2016-09-26
- Volume
- 34
- Issue
- 12
- Pages
- 1287-1291
- ISSN
- 0733-222X (ISSN-L)
- 1087-0156
- 1546-1696
- PMC
- PMC5143225
- OpenAlex
- W1897931464
- Cited by
- 186
- References
- 39
- Citations by year
- 2026: 10 · 2025: 13 · 2024: 15 · 2023: 9 · 2022: 38 · 2021: 29 · 2020: 18 · 2019: 20 · 2018: 19 · 2017: 10 · 2016: 4
- FWCI
- 9.18
- Type
- journal article
- Language
- en
- Open access
- green
- doi.org
- Licence
- Locations
- Nature Biotechnology (publishedVersion)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
- PubMed (publishedVersion)
- Europe PMC (PubMed Central) (acceptedVersion)
- PubMed Central (acceptedVersion)
- Digital Access to Scholarship at Harvard (DASH) (Harvard University) (submittedVersion, cc-by)
- Funders
- National Institutes of Health (NIH-R01, R01‐grant)
- Defense Advanced Research Projects Agency (DARPA-D12AP00025, D12AP00025)
- Advanced Research Projects Agency
- National Human Genome Research Institute (R01HG005220)
- National Cancer Institute (T32CA009337)
- National Institute of General Medical Sciences (R01GM083084)
- National Institute of Neurological Disorders and Stroke (R01NS054794)
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene expression and cancer classification
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
- Keywords
- RNA-seq, transcript abundance estimation, GC content bias, Alpine, false positives
- MeSH terms
- Algorithms, Base Composition, Computer Simulation, Models, Genetic, RNA, Sensitivity and Specificity, Software, Transcription Factors, Reproducibility of Results, Models, Statistical, Artifacts, Sequence Analysis, RNA
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 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
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
