Fast and accurate protein structure search with Foldseek
Open accessPublished version of a preprint
Abstract
As structure prediction methods are generating millions of publicly available protein structures, searching these databases is becoming a bottleneck. Foldseek aligns the structure of a query protein against a database by describing tertiary amino acid interactions within proteins as sequences over a structural alphabet. Foldseek decreases computation times by four to five orders of magnitude with 86%, 88% and 133% of the sensitivities of Dali, TM-align and CE, respectively.
AI snapshot
Written from the paper’s full text.
TL;DR
Foldseek encodes protein tertiary interactions as sequences over a 20-state 3Di structural alphabet and searches them with MMseqs2-derived prefilters and alignment modules. It reduces structure-search computation times by four to five orders of magnitude while retaining 86%, 88% and 133% of the sensitivities of Dali, TM-align and CE, respectively.
Key findings
- Foldseek decreases computation times by four to five orders of magnitude with 86%, 88% and 133% of the sensitivities of Dali, TM-align and CE, respectively.
- On SCOPe40 all-versus-all searches, Foldseek is more than 4,000 times faster than TM-align and Dali and over 21,000 times faster than CE; on AlphaFoldDB (version 1), it is around 184,600 and 23,000 times faster than Dali and TM-align, respectively.
- In precision-recall analysis, Foldseek-TM and Foldseek have the highest and third-highest area under the precision-recall curve on family, superfamily and fold levels; Foldseek-TM improves over TM-align because its prefilter suppresses high-scoring false positives.
- For SARS-CoV-2 RdRp (PDB: 6M71, chain A; 942 residues) searched in AlphaFoldDB (version 1), search times were 10 d for Dali, 33 h for TM-align and 6 s for Foldseek, making it 180,000 and 23,000 times faster; all top 10 hits were known RdRp homologs.
Methods
- Foldseek discretizes structures into 3Di sequences describing each residue’s geometric conformation with its spatially closest residue, then uses MMseqs2’s double-diagonal k-mer and gapless alignment prefilters followed by SIMD-accelerated Smith–Waterman local alignment combining 3Di and amino acid substitution scores; an optional global TM-align mode is also provided.
- Benchmarks include SCOPe40 (11,211 non-redundant single-domain structures) all-versus-all, AlphaFoldDB (version 1) clustered to 34,270 structures with 100 randomly selected queries for reference-free multi-domain tests, and HOMSTRAD (1,032 protein families); runtime evaluations used an AMD EPYC 7702P 64-core CPU.
- Sensitivity was measured as the fraction of true positives before the first false positive, with AUROC/ROC1 and precision-recall analyses; multi-domain benchmarks used LDDT thresholds of at least 0.6 for true positives and below 0.25 for false positives, ignoring matches in between.
Limitations
- The paper states that the SCOPe benchmark uses only single-domain proteins as queries and targets, so it cannot assess local similarities such as homologous domains shared between multi-domain proteins; the HOMSTRAD alignment benchmark has the same limitation, as most HOMSTRAD proteins are single-domain (median length 182 residues).
- The paper states that the reference-free multi-domain benchmark requires choosing thresholds for true and false positives.
- The text makes evident that the 3Di alphabet and substitution matrix were trained and cross-validated on SCOPe40, and that methods likely overfitted on SCOPe due to data leakage were excluded from benchmark comparison; the paper reports no overfitting despite training 492 parameters.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
0 of 22 cited sources checked · 0 supported · 0 flagged
- Findingown data
Among 133,813 high-scoring Foldseek false positives in AlphaFoldDB self-search, 1,675 were inspected and revealed queries with multiple structured segments but incorrect relative orientations.
“We inspected the 1,675 (of 133,813) high-scoring FPs (score per aligned column ≥ 1.0, TM-score < 0.5), revealing queries with multiple structured segments but with incorrect relative orientations (Supplementary Table 3 and Supplementary Fig. 9).”
Quote in the paperSubstantial claim - Findingown data
In the HOMSTRAD alignment quality benchmark, Foldseek performs slightly below CE, Dali and TM-align.
“In the reference-based HOMSTRAD alignment quality benchmark24, Foldseek performs slightly below CE, Dali and TM-align (Fig. 2e).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [24] Mizuguchi, K., Deane, C. M., Blundell, T. L. & Overington, J. P. HOMSTRAD: a database of protein structure alignments for homologous families. Protein Sci. 7, 2469–2471 (1998). source
Not checked in the free run.
- Findingown data
3Di has more evenly distributed state frequencies than traditional backbone structural alphabets.
“(1) Weaker dependency between consecutive letters and (2) more evenly distributed state frequencies, both enhancing information density and reducing false positives (FPs) (Supplementary Table 1).”
Quote in the paperSubstantial claim - Findingown data
Foldseek's performance is similar across all six secondary structure classes in SCOPe.
“Foldseek’s performance is similar across all six secondary structure classes in SCOPe (Supplementary Fig. 5).”
Quote in the paperNot a substantial claim - Findingown data
3Di has weaker dependency between consecutive letters than traditional backbone structural alphabets.
“(1) Weaker dependency between consecutive letters and (2) more evenly distributed state frequencies, both enhancing information density and reducing false positives (FPs) (Supplementary Table 1).”
Quote in the paperSubstantial claim
- Findingown data
The folded segments in those high-scoring false positives were correctly aligned by Foldseek.
“The folded segments were correctly aligned by Foldseek.”
Quote in the paperSubstantial claim - Findingown data
The 3D-BLAST matrix with Smith-Waterman resulted in a similar performance to CLE: 0.717, 0.230 and 0.011 for family classification, superfamily classification and fold classification, respectively.
“We noticed that the 3D-BLAST matrix with Smith–Waterman resulted in a similar performance to CLE: 0.717, 0.230 and 0.011 for family classification, superfamily classification and fold classification, respectively.”
Quote in the paperNot a substantial claim - Findingown data
All top 10 hits for SARS-CoV-2 RdRp were known RdRp homologs.
“All top 10 hits were known RdRp homologs (Supplementary Table 4).”
Quote in the paperSubstantial claim - Findingown data
No overfitting was observed for the 3Di alphabet despite training 492 parameters (282 neural network and 210 substitution matrix entries).
“No overfitting was observed, despite training 492 parameters (282 neural network and 210 substitution matrix entries).”
Quote in the paperNot a substantial claim - Findingown data
Dali, CE and CLE-Smith-Waterman failed to produce an alignment for 35, 1 and 1 out of 1,032 HOMSTRAD pairs, respectively, which were rated with a sensitivity of 0.
“Dali, CE and CLE-Smith–Waterman failed to produce an alignment for 35, 1 and 1 out of 1,032 pairs, respectively, which were rated with a sensitivity of 0.”
Quote in the paperNot a substantial claim - Findingown data
The average TM-score performed best in the single-domain benchmark.
“Our results showed that the average TM-score performed the best in our single-domain benchmark.”
Quote in the paperNot a substantial claim - Findingown data
The virtual center defined by θ = 270°, τ = 0° and l = 2 performed best in the SCOPe benchmark.
“The virtual center defined by θ = 270∘, τ = 0∘ and l = 2 performed best in the SCOPe benchmark.”
Quote in the paperSubstantial claim - Findingreasoning
The gapless double-match criterion suppresses hits to non-homologous structures effectively.
“The gapless double-match criterion suppresses hits to non-homologous structures effectively, as they are less likely to have consecutive k-mer matches on the same diagonal by chance.”
Quote in the paperSubstantial claim - Premisecitation
No similar prefilters exist for structure alignment as exist for sequence search.
“First, whereas sequence search tools employ fast and sensitive prefilter algorithms to gain orders of magnitude in speed, no similar prefilters exist for structure alignment.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [12] Hasegawa, H. & Holm, L. Advances and pitfalls of protein structural alignment. Curr. Opin. Struct. Biol. 19, 341–348 (2009). source
Not checked in the free run.
- Interpretationown data
Foldseek-TM improves over TM-align because its prefilter suppresses high-scoring false positives.
“Notably, Foldseek-TM improves over TM-align because its prefilter suppresses high-scoring FPs.”
Quote in the paperSubstantial claim - Premisecitation
Detecting similarity between protein structures by three-dimensional superposition offers higher sensitivity for identifying homologous proteins.
“Detecting similarity between protein structures by three-dimensional (3D) superposition offers higher sensitivity for identifying homologous proteins10.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [10] Illergård, K., Ardell, D. H. & Elofsson, A. Structure is three to ten times more conserved than sequence—a study of structural response in protein cores. Proteins 77, 499–508 (2009). source
Not checked in the free run.
- Premisecitation
A crucial idea to increase speed is to describe the amino acid backbone of proteins as sequences over a structural alphabet and compare structures using sequence alignments.
“To increase speed, a crucial idea is to describe the amino acid backbone of proteins as sequences over a structural alphabet and compare structures using sequence alignments15.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [15] Guyon, F., Camproux, A.-C., Hochez, J. & Tuffery, P. SA-Search: a web tool for protein structure mining based on a structural alphabet. Nucleic Acids Res. 32, W545–W548 (2004). source
Not checked in the free run.
- Premisecitation
Structural alphabets reduce structure comparisons to much faster sequence alignments.
“Structural alphabets thus reduce structure comparisons to much faster sequence alignments.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [15] Guyon, F., Camproux, A.-C., Hochez, J. & Tuffery, P. SA-Search: a web tool for protein structure mining based on a structural alphabet. Nucleic Acids Res. 32, W545–W548 (2004). source
Not checked in the free run.
- Premisecitation
Many proteins cannot be annotated because detecting distant evolutionary relationships from sequences alone remains challenging.
“many proteins cannot be annotated because detecting distant evolutionary relationships from sequences alone remains challenging9.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [9] Mahlich, Y., Steinegger, M., Rost, B. & Bromberg, Y. HFSP: high speed homology-driven function annotation of proteins. Bioinformatics 34, i304–i312 (2018). source
Not checked in the free run.
- Premisecitation
Most structural aligners, such as TM-align, Dali and CE, solve the alignment optimization problem by iterative or stochastic optimization.
“Most structural aligners, such as the popular TM-align, Dali and CE11,13,14, solve the alignment optimization problem by iterative or stochastic optimization.”
Quote in the paperSubstantial claim0 of 3 sources checked- cites [11] Zhang, Y. & Skolnick, J. TM-align: a protein structure alignment algorithm based on the TM-score. Nucleic Acids Res. 33, 2302–2309 (2005). source
Not checked in the free run.
- cites [13] Holm, L. Using Dali for protein structure comparison. Methods Mol. Biol. 2112, 29–42 (2020). source
Not checked in the free run.
- cites [14] Shindyalov, I. N. & Bourne, P. E. Protein structure alignment by incremental combinatorial extension (CE) of the optimal path. Protein Eng. 11, 739–747 (1998). source
Not checked in the free run.
- Premisecitation
Structural similarity scores are non-local: changing the alignment in one part affects the similarity in all other parts.
“Second, structural similarity scores are non-local: changing the alignment in one part affects the similarity in all other parts.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [12] Hasegawa, H. & Holm, L. Advances and pitfalls of protein structural alignment. Curr. Opin. Struct. Biol. 19, 341–348 (2009). source
Not checked in the free run.
- Study design and conductreasoning
Methods that trained many thousands of parameters with strong data leakage among training, validation and test sets are likely to be overfitted on SCOPe and were excluded from the benchmark.
“Owing to dataset overlap, we excluded methods from the benchmark that are likely to be overfitted on SCOPe. This applies to methods that trained many thousands of parameters (for example, deep neural networks) with strong data leakage among training, validation and test sets.”
Quote in the paperSubstantial claim - Interpretationown data
The HOMSTRAD alignment benchmark has the same limitation as the SCOPe benchmark because the vast majority of proteins in HOMSTRAD have a single domain (median length = 182 residues).
“The alignment benchmark based on HOMSTRAD (Fig. 2e) has the same limitation, as the vast majority of proteins in HOMSTRAD have a single domain (median length = 182 residues).”
Quote in the paperNot a substantial claim - Interpretationreasoning
The LDDT score function does not properly penalize non-compact models in which each residue has few neighbors within 15 Å.
“This does not properly penalize non-compact models in which each residue has few neighbors within 15 Å.”
Quote in the paperSubstantial claim - Interpretationreasoning
Foldseek enables fast and sensitive comparison of large structure sets.
“Foldseek enables fast and sensitive comparison of large structure sets.”
Quote in the paperSubstantial claim - Interpretationreasoning
The SCOPe benchmark cannot assess the ability of tools to find local similarities, such as homologous domains shared between two multi-domain proteins, because it uses only single-domain proteins as queries and targets.
“The SCOPe benchmark to measure search sensitivity uses only single-domain proteins as queries and targets (Fig. 2a–c). It, therefore, cannot assess the ability of tools to find local similarities—for example, finding homologous domains shared between two multi-domain proteins.”
Quote in the paperSubstantial claim - Interpretationreasoning
When the virtual center falls back to i + 1 or i − 1, the interaction captures only the backbone conformation.
“In that case, the interaction captures only the backbone conformation.”
Quote in the paperNot a substantial claim - Study design and conductown data
The 3Di alphabet and substitution matrix were trained by four-fold cross-validation on SCOPe40.
“We trained the 3Di alphabet (the VQ-VAE weights) and the substitution matrix by four-fold cross-validation on SCOPe40.”
Quote in the paperNot a substantial claim - Interpretationown data
The main limitation of structure comparisons-the four orders of magnitude slower speed-is removed by Foldseek.
“The main limitation in our view—the four orders of magnitude slower speed of structure comparisons—is removed by Foldseek.”
Quote in the paperSubstantial claim - Interpretationreasoning
Using MMseqs2's pre-computed database indexing capabilities in Foldseek allows search databases to be fully cached in system memory and instantly accessed by each Foldseek process, avoiding expensive disk accesses.
“Using these, the search databases can be fully cached in system memory by the operating system and instantly accessed by each Foldseek process, thus avoiding expensive accesses to slow disk drives.”
Quote in the paperSubstantial claim - Premisecitation
The HOMSTRAD database contains expert-curated homologous structural alignments for 1,032 protein families.
“The HOMSTRAD database contains expert-curated homologous structural alignments for 1,032 protein families24.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [24] Mizuguchi, K., Deane, C. M., Blundell, T. L. & Overington, J. P. HOMSTRAD: a database of protein structure alignments for homologous families. Protein Sci. 7, 2469–2471 (1998). source
Not checked in the free run.
- May combine two claims. 22
- Classified as probably not a substantial claim. 17
May combine two claims.
Findingnone givenFoldseek decreases computation times by four to five orders of magnitude with 86%, 88% and 133% of the sensitivities of Dali, TM-align and CE, respectively.
“Foldseek decreases computation times by four to five orders of magnitude with 86%, 88% and 133% of the sensitivities of Dali, TM-align and CE, respectively.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataOn the SCOPe40 benchmark set, Foldseek is more than 4,000 times faster than TM-align and Dali and over 21,000 times faster than CE.
“On this small SCOPe40 benchmark set, Foldseek is more than 4,000 times faster than TM-align and Dali and over 21,000 times faster than CE (Fig. 2c).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataFoldseek alignments are more accurate and sensitive than MMseqs2, CLE-SW and TM-align, similarly accurate as Dali and 13% less precise but 15% more sensitive than CE.
“Foldseek alignments are more accurate and sensitive than MMseqs2, CLE-SW and TM-align, similarly accurate as Dali and 13% less precise but 15% more sensitive than CE.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataOn AlphaFoldDB (version 1), Foldseek is around 184,600 and 23,000 times faster than Dali and TM-align, respectively.
“On the much larger AlphaFoldDB (version 1), where Foldseek approaches its full speed, it is around 184,600 and 23,000 times faster than Dali and TM-align, respectively (see below).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataOn the full-length multi-domain AlphaFoldDB benchmark, Foldseek has similar sensitivity as Dali, CE and TM-align and much higher sensitivity than CLE-SW and MMseqs2.
“Again, Foldseek has similar sensitivity as Dali, CE and TM-align and much higher sensitivity than CLE-SW and MMseqs2.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataIn precision-recall analysis, Foldseek-TM and Foldseek have the highest and third-highest area under the precision-recall curve on each of the three levels.
“In a precision-recall analysis, Foldseek-TM and Foldseek have the highest and third-highest area under the precision-recall curve on each of the three levels (Fig. 2b and Supplementary Fig. 4).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataFoldseek's sensitivity is below Dali and TM-align, higher than CE and much above 3D-BLAST and CLE-SW.
“Foldseek’s sensitivity is below Dali and TM-align, higher than the structural aligner CE and much above the structural alphabet-based search tools 3D-BLAST and CLE-SW (Fig. 2a).”
Quote in the paperSubstantial claimMay combine two claims.
Findingnone givenThe highest information density in 3Di is encoded in conserved protein cores and the lowest in non-conserved coil/loop regions, whereas the opposite is true for backbone structural alphabets.
“(3) The highest information density is encoded in conserved protein cores and the lowest in non-conserved coil/loop regions, whereas the opposite is true for backbone structural alphabets.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Findingown dataDali's conversion to DAT format produced 11,137 valid structures out of the 11,211 initial structures for the SCOPe benchmark and 34,252 structures out of 34,270 SPICi clusters.
“The conversion to DAT format produced 11,137 valid structures out of the 11,211 initial structures for the SCOPe benchmark and 34,252 structures out of 34,270 SPICi clusters.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown data3D-BLAST produced occasionally high scores (>107) for single residue alignments, so its measurement was excluded from the multi-domain benchmark.
“We excluded 3D-BLAST’s measurement from the multi-domain benchmark because it produced occasionally high scores (>107) for single residue alignments.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Findingnone givenFoldseek includes four variants of AlphaFoldDB version 4: UniProt (214 million structures), UniProt50 (54 million structures), Proteome (564,000 structures) and Swiss-Prot (542,000 structures).
“Currently, we include the four variants of the AlphaFoldDB (version 4): UniProt (214 million structures), UniProt50, a clustered database to 50% sequence identity and 90% bi-directional coverage using MMseqs2 (parameters -c 0.9 --min-seq-id 0.5 --cluster-reassign; 54 million structures), Proteome (564,000 structures) and Swiss-Prot (542,000 structures).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingnone givenDuring batch searches, Foldseek adapts its memory use to the available RAM of the machine.
“During batch searches, Foldseek adapts its memory use to the available RAM of the machine.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataFoldseek finds around the same number of false positives and true positives with a score of 51 in SCOPe40, so the probability for a hit with score 51 is 50%.
“For example, Foldseek finds around the same number of FPs and TPs with a score of 51 in SCOPe40. The probability for a hit with score 51 is, therefore, 50%.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataThe HOMSTRAD pairwise alignments between the first and last members of each family resulted in structures of a median length of 182 residues.
“We downloaded the latest HOMSTRAD version (https://mizuguchilab.org/homstrad/data/homstrad_with_PDB_2022_Aug_1.tar.gz) and picked the pairwise alignments between the first and last members of each family, which resulted in structures of a median length of 182 residues.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataIn a SARS-CoV-2 RdRp search in AlphaFoldDB, search times were 10 d for Dali, 33 h for TM-align and 6 s for Foldseek, making Foldseek 180,000 and 23,000 times faster.
“Search times were 10 d for Dali, 33 h for TM-align and 6 s for Foldseek, making it 180,000 and 23,000 times faster.”
Quote in the paperSubstantial claimMay combine two claims.
Findingnone givenThe 3Di alphabet developed for Foldseek describes tertiary residue-residue interactions instead of backbone conformations and proved critical for reaching high sensitivities.
“The 3Di alphabet developed for Foldseek describes tertiary residue–residue interactions instead of backbone conformations and proved critical for reaching high sensitivities.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataOne query in the multi-domain benchmark was excluded because it ran over 16 days.
“For the multi-domain benchmark, we excluded one query that was running over 16 d.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataThe SCOPe benchmark set consists of single domains with an average length of 174 residues.
“The SCOPe benchmark set consists of single domains with an average length of 174 residues.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Findingnone givenThe AlphaFoldDB/UniProt50 (version 4), AlphaFoldDB/Proteome (version 4), AlphaFoldDB/Swiss-Prot (version 4), CATH50, ESM Atlas High-Quality Clu30 and PDB100 require 191 GB, 3.8 GB, 3.4 GB, 1.4 GB, 110 GB and 2.0 GB RAM, respectively.
“The AlpaFoldDB/UniProt50 (version 4), AlphaFoldDB/Proteome (version 4), AlphaFoldDB/Swiss-Prot (version 4), CATH50, ESM Atlas High-Quality Clu30 and PDB100 require 191 GB, 3.8 GB, 3.4 GB, 1.4 GB, 110 GB and 2.0 GB RAM, respectively.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFoldseek's Needleman-Wunsch implementation does not compute exactly the same alignment as TM-align because it applies affine gap costs, whereas TM-align does not.
“Our NW implementation does not compute exactly the same alignment because we apply affine gap costs, whereas TM-align does not (Supplementary Fig. 12).”
Quote in the paperNot a substantial claimMay combine two claims.
FindingreasoningBy not restricting itself to exact matches, Foldseek's prefilter achieves high sensitivity while reducing the number of sequences for which full alignments are computed by several orders of magnitude.
“By not restricting itself to exact matches, the prefilter achieves high sensitivity while reducing the number of sequences for which full alignments are computed by several orders of magnitude.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataSPICi produced 34,270 clusters from the all-against-all BLAST search of AlphaFoldDB version 1.
“SPICi produced 34,270 clusters from the search result.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
PremisecitationThe European Bioinformatics Institute holds over 214 million structures predicted by AlphaFold2.
“The European Bioinformatics Institute already holds over 214 million structures predicted by AlphaFold2 (ref. 3)”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [3] Varadi, M. et al. AlphaFold Protein Structure Database: massively expanding the structural coverage of protein–sequence space with high-accuracy models. Nucleic Acids Res. 50, D439–D444 (2022). source
Not checked in the free run.
May combine two claims.
PremisecitationSequence searching is four to five orders of magnitude faster: an all-versus-all comparison of 100 million sequences would take MMseqs2 only around a week on the same cluster.
“Sequence searching is four to five orders of magnitude faster: an all-versus-all comparison of 100 million sequences would take MMseqs2 (ref. 6) only around a week on the same cluster.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [6] Steinegger, M. & Söding, J. MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets. Nat. Biotechnol. 35, 1026–1028 (2017). source
Not checked in the free run.
May combine two claims.
PremisecitationSearching with a single query structure through a database with 100 million protein structures would take TM-align a month on one CPU core, and an all-versus-all comparison would take 10 millennia on a 1,000-core cluster.
“Searching with a single query structure through a database with 100 million protein structures would take the popular TM-align11 tool a month on one CPU core, and an all-versus-all comparison would take 10 millennia on a 1,000-core cluster.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [11] Zhang, Y. & Skolnick, J. TM-align: a protein structure alignment algorithm based on the TM-score. Nucleic Acids Res. 33, 2302–2309 (2005). source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe ESM Atlas contains over 617 million metagenomic structures predicted by ESMFold.
“the ESM Atlas contains over 617 million metagenomic structures predicted by ESMFold4.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [4] Lin, Z. et al. Evolutionary-scale prediction of atomic-level protein structure with a language model. Science 379, 1123–1130 (2023). source
Not checked in the free run.
May combine two claims.
Interpretationown data3Di's weaker dependency and more evenly distributed state frequencies enhance information density and reduce false positives.
“(1) Weaker dependency between consecutive letters and (2) more evenly distributed state frequencies, both enhancing information density and reducing false positives (FPs) (Supplementary Table 1).”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationMost structural alphabets, such as CLE, 3D-BLAST and Protein Blocks, discretize the conformations of short stretches of usually 3-5 Cα atoms.
“Most, such as CLE, 3D-BLAST and Protein Blocks, discretize the conformations of short stretches of usually 3–5 Cα atoms17–19.”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [17] Wang, S. & Zheng, W.-M. CLePAPS: fast pair alignment of protein structures based on conformational letters. J. Bioinform. Comput. Biol. 6, 347–366 (2008). source
Not checked in the free run.
- cites [18] Yang, J.-M. & Tung, C.-H. Protein structure database search and evolutionary classification. Nucleic Acids Res. 34, 3646–3659 (2006). source
Not checked in the free run.
- cites [19] de Brevern, A. G., Etchebest, C. & Hazout, S. Bayesian probabilistic approach for predicting backbone structures in terms of protein blocks. Proteins 41, 271–287 (2000). source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
InterpretationreasoningDatabases are only required to remain resident in RAM if Foldseek is used as a webserver.
“Databases are only required to remain resident in RAM if Foldseek is used as a webserver.”
Quote in the paperNot a substantial claimMay combine two claims.
Interpretationown data3D aligners such as TM-align may overlook homologous structures that are not globally superposable, whereas Foldseek and Dali are independent of relative domain orientations and excel at detecting homologous multi-domain structures.
“This illustrates that 3D aligners such as TM-align may overlook homologous structures that are not globally superposable, whereas Foldseek (as well as the two-dimensional (2D) aligner Dali) is independent of relative domain orientations and excels at detecting homologous multi-domain structures12.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [12] Hasegawa, H. & Holm, L. Advances and pitfalls of protein structural alignment. Curr. Opin. Struct. Biol. 19, 341–348 (2009). source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
Interpretationnone givenA drawback of the reference-free multi-domain benchmark is the need to choose thresholds for true positives and false positives.
“A drawback of our reference-free multi-domain benchmark is the need to choose thresholds for TPs and FPs (Supplementary Fig. 6).”
Quote in the paperNot a substantial claimMay combine two claims.
Premisenone givenGeometricus is a representative of alignment-free tools, which are fast but can find only globally similar structures.
“We included Geometricus20 in the SCOPe benchmark as a representative of alignment-free tools, which are fast but can find only globally similar structures.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [20] Durairaj, J., Akdel, M., de Ridder, D. & van Dijk, A. D. J. Geometricus represents protein structures as shape-mers derived from moment invariants. Bioinformatics 36, i718–i725 (2020). source
Not checked in the free run.
May combine two claims.
Interpretationown dataThe virtual center preferably selects long-range, tertiary interactions and only falls back to selecting interactions to i + 1 or i − 1 when no other residues are nearby.
“This virtual center preferably selects long-range, tertiary interactions and only falls back to selecting interactions to i + 1 or i − 1 when no other residues are nearby.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S4 (based).
Interpretationnone givenMeasuring sensitivity up to the 1st false positive (ROC1) better reflects the requirements for low false discovery rates in automatic searches than measuring up to the 5th false positive.
“We measure the sensitivity up to the 1st FP (ROC1) instead, for example, up to the 5th FP, because ROC1 better reflects the requirements for low false discovery rates in automatic searches.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
InterpretationcitationWeighting true positive, false positive and false negative counts by the reciprocal of their family, superfamily or fold size makes folds, superfamilies and families contribute linearly with their size instead of quadratically.
“All counts (TP, FP and FN) were weighted by the reciprocal of their family, superfamily or fold size. In this way, folds, superfamilies and families contribute linearly with their size instead of quadratically36.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [36] Eddy, S. R. Accelerated profile HMM searches. PLoS Comput. Biol. 7, e1002195 (2011). source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe Foldseek webserver is based on the MMseqs2 webserver.
“The Foldseek webserver is based on the MMseqs2 webserver43.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [43] Mirdita, M., Steinegger, M. & Söding, J. MMseqs2 desktop and local web server app for fast, interactive sequence searches. Bioinformatics 35, 2856–2858 (2019). source
Not checked in the free run.
The paper relies on this without support.
The SCOPe40 dataset is a suitable benchmark for evaluating protein structure search sensitivity.
“The SCOPe benchmark set consists of single domains with an average length of 174 residues.”
The paper relies on this without support.
The HOMSTRAD pairwise alignments are accurate reference alignments for measuring alignment quality.
“With the HOMSTRAD alignments as reference, we measured for each pairwise alignment the sensitivity (fraction of residue pairs of the HOMSTRAD alignment that were correctly aligned) and the precision (fraction of correctly aligned residue pairs in the predicted alignment).”
The paper relies on this without support.
The ESM Atlas API can accurately predict structures of user-supplied sequences up to 400 residues.
“We use the ESM Atlas API to predict structures of user-supplied sequences that are, at most, 400 residues long.”
The paper relies on this without support.
The virtual center parameters optimized on SCOPe generalize to other datasets.
“The virtual center defined by θ = 270∘, τ = 0∘ and l = 2 performed best in the SCOPe benchmark.”
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?
Foldseek increases the speed of structural comparisons by four to five orders of magnitude relative to previous approaches while maintaining sensitivity.
“Foldseek can increase the speed of comparisons of structures by four to five orders of magnitudes relative to previous approaches while maintaining sensitivity15”
From Clustering predicted structures at the scale of the known protein universe (2023) · cites this paper as [15]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Same-family hits are both included as TPs and ignored
MinorA hit within the same family is also within the same superfamily, so the stated TP and ignored categories overlap.
“Here, hits within the same superfamily are TP;”
“hits to the same family or to another superfamily are ignored.”
Conflicting descriptions of the E-value network's input composition
MinorThe main text says the network uses 3Di sequence composition, whereas Methods says it uses amino acid composition. These are different inputs for the same predicted distribution parameters.
“E values are calculated using an extreme-value score distribution, with parameters predicted by a neural network based on 3Di sequence composition and query length”
“We trained a neural network to predict μ and λ from the amino acid composition of the query and its length (so a scrambled version of the query sequence would produce the same μ and λ).”
Cites
Basic local alignment search tool
Altschul, S. F., Gish, W., Miller, W., Myers, E. W., & Lipman, D. J. (1990) · Journal of Molecular Biology · doi:10.1006/jmbi.1990.9999 · Cited by 98,669 · see citing works
Highly accurate protein structure prediction with AlphaFold
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., et al. (2021) · Nature · doi:10.1038/s41586-021-03819-2 · Cited by 43,764 · see citing works
Self-citationVaradi, M., Anyango, S., Deshpande, M., Nair, S., Natassia, C., Yordanova, G., Yuan, D., Stroe, O., Wood, G., Laydon, A., et al. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkab1061 · Cited by 7,850 · see citing works
Accelerated Profile HMM Searches
Eddy, S. R. (2011) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1002195 · Cited by 6,975 · see citing works
Accurate prediction of protein structures and interactions using a 3-track network
Baek, M., DiMaio, F., Anishchenko, I., Dauparas, J., Ovchinnikov, S., Lee, G. R., Wang, J., Cong, Q., Kinch, L. N., Schaeffer, R. D., et al. (2021) · doi:10.1101/2021.06.14.448402 · Cited by 5,542 · see citing works
Evolutionary-scale prediction of atomic level protein structure with a language model
Lin, Z., Akin, H., Rao, R., Hie, B., Zhu, Z., Lu, W., Smetanin, N., Verkuil, R., Kabeli, O., Shmueli, Y., et al. (2022) · doi:10.1101/2022.07.20.500902 · Cited by 5,523 · see citing works
Sensitive protein alignments at tree-of-life scale using DIAMOND
Buchfink, B., Reuter, K., & Drost, H.-G. (2021) · Nature Methods · doi:10.1038/s41592-021-01101-x · Cited by 5,097 · see citing works
MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets
Steinegger, M., & Söding, J. (2017) · Nature Biotechnology · doi:10.1038/nbt.3988 · Cited by 4,840 · see citing works
Self-citationTM-align: a protein structure alignment algorithm based on the TM-score
Zhang, Y. (2005) · Nucleic Acids Research · doi:10.1093/nar/gki524 · Cited by 3,774 · see citing works
Protein structure alignment by incremental combinatorial extension (CE) of the optimal path.
Shindyalov, I., & Bourne, P. (1998) · Protein Engineering · doi:10.1093/protein/11.9.739 · Cited by 2,199 · see citing works
Why are ?natively unfolded? proteins unstructured under physiologic conditions?
Uversky, V. N., Gillespie, J. R., & Fink, A. L. (2000) · Proteins Structure Function and Bioinformatics · doi:10.1002/1097-0134(20001115)41:3 · Cited by 2,063 · see citing works
Structure of the RNA-dependent RNA polymerase from COVID-19 virus
Gao, Y., Yan, L., Huang, Y., Liu, F., Zhao, Y., Cao, L., Wang, T., Sun, Q., Ming, Z., Zhang, L., et al. (2020) · Science · doi:10.1126/science.abb7498 · Cited by 1,614 · see citing works
Burley, S. K., Bhikadiya, C., Bi, C., Bittrich, S., Chen, L., Crichlow, G. V., Christie, C. H., Dalenberg, K., Di Costanzo, L., Duarte, J. M., et al. (2021) · Nucleic Acids Research · doi:10.1093/nar/gkaa1038 · Cited by 1,581 · see citing works
mixtools : An R Package for Analyzing Finite Mixture Models
Benaglia, T., Chauveau, D., Hunter, D. R., & Young, D. S. (2009) · Journal of Statistical Software · doi:10.18637/jss.v032.i06 · Cited by 1,351 · see citing works
Burley, S. K., Berman, H. M., Bhikadiya, C., Bi, C., Chen, L., Costanzo, L. D., Christie, C. H., Dalenberg, K., Duarte, J. M., Dutta, S., et al. (2018) · Nucleic Acids Research · doi:10.1093/nar/gky1004 · Cited by 1,249 · see citing works
Protein Data Bank: the single global archive for 3D macromolecular structure data
wwPDB consortium, Burley, S. K., Berman, H. M., Bhikadiya, C., Bi, C., Chen, L., Costanzo, L. D., Christie, C., Duarte, J. M., Dutta, S., et al. (2019) · Nucleic Acids Research · doi:10.1093/nar/gky949 · Cited by 1,163 · see citing works
HH-suite3 for fast remote homology detection and deep protein annotation
Steinegger, M., Meier, M., Mirdita, M., Vöhringer, H., Haunsberger, S. J., & Söding, J. (2019) · BMC Bioinformatics · doi:10.1186/s12859-019-3019-7 · Cited by 1,145 · see citing works
Self-citationMariani, V., Biasini, M., Barbato, A., & Schwede, T. (2013) · Bioinformatics · doi:10.1093/bioinformatics/btt473 · Cited by 1,138 · see citing works
MMseqs2 desktop and local web server app for fast, interactive sequence searches
Mirdita, M., Steinegger, M., & Söding, J. (2019) · Bioinformatics · doi:10.1093/bioinformatics/bty1057 · Cited by 618 · see citing works
Self-citationUsing Dali for Protein Structure Comparison
Holm, L. (2020) · Methods in molecular biology · doi:10.1007/978-1-0716-0270-6_3 · Cited by 613 · see citing works
Illergård, K., Ardell, D. H., & Elofsson, A. (2009) · Proteins Structure Function and Bioinformatics · doi:10.1002/prot.22458 · Cited by 559 · see citing works
HOMSTRAD: A database of protein structure alignments for homologous families
Mizuguchi, K., Deane, C. M., Blundell, T. L., & Overington, J. P. (1998) · Protein Science · doi:10.1002/pro.5560071126 · Cited by 501 · see citing works
NGL viewer: web-based molecular graphics for large complexes
Rose, A. S., Bradley, A. R., Valasatava, Y., Duarte, J. M., Prlić, A., & Rose, P. W. (2018) · Bioinformatics · doi:10.1093/bioinformatics/bty419 · Cited by 489 · see citing works
Striped Smith–Waterman speeds database searches six times over other SIMD implementations
Farrar, M. S. (2006) · Bioinformatics · doi:10.1093/bioinformatics/btl582 · Cited by 417 · see citing works
Fast procedure for reconstruction of full‐atom protein models from reduced representations
Rotkiewicz, P., & Skolnick, J. (2008) · Journal of Computational Chemistry · doi:10.1002/jcc.20906 · Cited by 395 · see citing works
Advances and pitfalls of protein structural alignment
Hasegawa, H., & Holm, L. (2009) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2009.04.003 · Cited by 369 · see citing works
MMseqs software suite for fast and deep clustering and searching of large protein sequence sets
Hauser, M., Steinegger, M., & Söding, J. (2016) · Bioinform. · doi:10.1093/bioinformatics/btw006 · Cited by 265 · see citing works
Self-citationParasail: SIMD C library for global, semi-global, and local pairwise sequence alignments
Daily, J. A. (2016) · BMC Bioinformatics · doi:10.1186/s12859-016-0930-z · Cited by 241 · see citing works
SSW Library: An SIMD Smith-Waterman C/C++ Library for Use in Genomic Applications
Zhao, M., Lee, W.-P., Garrison, E. P., & Marth, G. (2013) · PLoS ONE · doi:10.1371/journal.pone.0082138 · Cited by 239 · see citing works
SPICi: a fast clustering algorithm for large biological networks
Jiang, P., & Singh, M. (2010) · Bioinformatics · doi:10.1093/bioinformatics/btq078 · Cited by 239 · see citing works
Chandonia, J., Fox, N. K., & Brenner, S. E. (2018) · Nucleic Acids Research · doi:10.1093/nar/gky1134 · Cited by 167 · see citing works
Protein structure database search and evolutionary classification
Yang, J. (2006) · Nucleic Acids Research · doi:10.1093/nar/gkl395 · Cited by 115 · see citing works
AlphaFold2 reveals commonalities and novelties in protein structure space for 21 model organisms
Bordin, N., Sillitoe, I., Nallapareddy, V., Rauer, C., Lam, S. D., Waman, V. P., Sen, N., Heinzinger, M., Littmann, M., Kim, S., et al. (2022) · doi:10.1101/2022.06.02.494367 · Cited by 103 · see citing works
Self-citationAlphaFold2 reveals commonalities and novelties in protein structure space for 21 model organisms
Bordin, N., Sillitoe, I., Nallapareddy, V., Rauer, C., Lam, S. D., Waman, V. P., Sen, N., Heinzinger, M., Littmann, M., Kim, S., et al. (2023) · Communications Biology · doi:10.1038/s42003-023-04488-9 · Cited by 103 · see citing works
Self-citationHH-suite3 for fast remote homology detection and deep protein annotationPreprint
Steinegger, M., Meier, M., Mirdita, M., et al. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/560029 · Cited by 89 · see citing works
Self-citationProtein sequence comparison and fold recognition: progress and good-practice benchmarking
Söding, J., & Remmert, M. (2011) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2011.03.005 · Cited by 80 · see citing works
Self-citationAlgorithms, Applications, and Challenges of Protein Structure Alignment
Ma, J., & Wang, S. (2014) · Advances in protein chemistry and structural biology · doi:10.1016/b978-0-12-800168-4.00005-6 · Cited by 59 · see citing works
Early ribbon drawings of proteins.
Richardson, J. S. (2000) · Nature Structural Biology · doi:10.1038/77912 · Cited by 52 · see citing works
CLePAPS: FAST PAIR ALIGNMENT OF PROTEIN STRUCTURES BASED ON CONFORMATIONAL LETTERS
Wang, S., & Zheng, W. (2008) · Journal of Bioinformatics and Computational Biology · doi:10.1142/s0219720008003461 · Cited by 51 · see citing works
SA-Search: a web tool for protein structure mining based on a Structural Alphabet
Guyon, F., Camproux, A., Hochez, J., & Tufféry, P. (2004) · Nucleic Acids Research · doi:10.1093/nar/gkh467 · Cited by 49 · see citing works
Geometricus represents protein structures as shape-mers derived from moment invariants
Durairaj, J., Akdel, M., de Ridder, D., & van Dijk, A. D. J. (2020) · Bioinformatics · doi:10.1093/bioinformatics/btaa839 · Cited by 47 · see citing works
BioJava 5: A community driven open-source bioinformatics library
Lafita, A., Bliven, S. E., Prlić, A., Guzenko, D., Rose, P. W., Bradley, A. R., Pavan, P., Myers-Turnbull, D., Valasatava, Y., Heuer, M., et al. (2019) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1006791 · Cited by 44 · see citing works
HFSP: high speed homology-driven function annotation of proteins
Mahlich, Y., Steinegger, M., Rost, B., & Bromberg, Y. (2018) · Bioinformatics · doi:10.1093/bioinformatics/bty262 · Cited by 42 · see citing works
Self-citationFoldcomp: a library and format for compressing and indexing large protein structure sets
Kim, H., Mirdita, M., & Steinegger, M. (2023) · Bioinformatics · doi:10.1093/bioinformatics/btad153 · Cited by 29 · see citing works
Self-citationDeepAlign, a 3D alignment method based on regionalized deep learning for Cryo-EM
Jiménez-Moreno, A., Střelák, D., Filipovič, J., Carazo, J. M., & Sorzano, C. Ó. S. (2021) · Journal of Structural Biology · doi:10.1016/j.jsb.2021.107712 · Cited by 23 · see citing works
Towards an efficient compression of 3D coordinates of macromolecular structures
Valasatava, Y., Bradley, A. R., Rose, A., Duarte, J. M., Prlić, A., & Rose, P. W. (2017) · PLoS ONE · doi:10.1371/journal.pone.0174846 · Cited by 14 · see citing works
fast_protein_cluster: parallel and optimized clustering of large-scale protein modeling data
Hung, L., & Samudrala, R. (2014) · Bioinformatics · doi:10.1093/bioinformatics/btu098 · Cited by 12 · see citing works
A cognitive organization theory (COT) of organizational change
van den Oord, A., Elliott, K., van Witteloostuijn, A. C., Barlage, M., Polós, L., & Rogiest, S. (2017) · Journal of Organizational Change Management · doi:10.1108/jocm-08-2016-0164 · Cited by 6 · see citing works
Flagged first, then by citations
Research works printed but not linked: 1
- Van den Oord, A., Vinyals, O. & Kavukcuoglu, K. Neural discrete representation learning. Proc. of the 31st Conference on Neural Information Processing Systems. https://proceedings.neurips.cc/paper_files/paper/2017/file/7a98af17e63a0ac09ce2e96d03992fbc-Paper.pdf (NIPS, 2017).
Cited by
Sorted by citations
Evolutionary-scale prediction of atomic level protein structure with a language model
Lin, Z., Akin, H., Rao, R., Hie, B., Zhu, Z., Lu, W., Smetanin, N., Verkuil, R., Kabeli, O., Shmueli, Y., et al. (2022) · doi:10.1101/2022.07.20.500902 · Cited by 5,523 · see citing works
InterPro: the protein sequence classification resource in 2025
Blum, M., Andreeva, A. E., Florentino, L. C., Chuguransky, S., Grego, T., Hobbs, E., Pinto, B. L., Orr, A., Paysan‐Lafosse, T., Ponamareva, I., et al. (2024) · Nucleic Acids Research · doi:10.1093/nar/gkae1082 · Cited by 1,180 · see citing works
Simulating 500 million years of evolution with a language model
Hayes, T., Rao, R., Akin, H., Sofroniew, N. J., Oktay, D., Lin, Z., Verkuil, R., Tran, V. Q., Deaton, J., Wiggert, M., et al. (2025) · Science · doi:10.1126/science.ads0018 · Cited by 937 · see citing works
Dali server: structural unification of protein families
Holm, L. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkac387 · Cited by 913 · see citing works
ProtGPT2 is a deep unsupervised language model for protein design
Ferruz, N., Schmidt, S., & Höcker, B. (2022) · Nature Communications · doi:10.1038/s41467-022-32007-7 · Cited by 795 · see citing works
DALI shines a light on remote homologs: One hundred discoveries
Holm, L., Laiho, A., Törönen, P., & Salgado, M. G. (2022) · Protein Science · doi:10.1002/pro.4519 · Cited by 685 · see citing works
ProGen2: Exploring the boundaries of protein language models
Nijkamp, E., Ruffolo, J. A., Weinstein, E. N., Naik, N., & Madani, A. (2023) · Cell Systems · doi:10.1016/j.cels.2023.10.002 · Cited by 480 · see citing works
Database resources of the National Center for Biotechnology Information in 2023
Sayers, E. W., Bolton, E., Brister, J. R., Canese, K., Chan, J., Comeau, D. C., Farrell, C. M., Feldgarden, M., Fine, A. M., Funk, K., et al. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkac1032 · Cited by 436 · see citing works
Self-citationLearning inverse folding from millions of predicted structuresPreprint
Hsu, C., Verkuil, R., Liu, J. J., Lin, Z., Hie, B., Sercu, T., Lerer, A. K., & Rives, A. W. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.04.10.487779 · Cited by 434 · see citing works
The Pfam protein families database: embracing AI/ML
Paysan-Lafosse, T., Andreeva, A., Blum, M., Chuguransky, S., Grego, T., Pinto, B. L., Salazar, G. A., Bileschi, M. L., Llinares-López, F., Meng-Papaxanthos, L., et al. (2024) · Nucleic Acids Research · doi:10.1093/nar/gkae997 · Cited by 382 · see citing works
One-shot design of functional protein binders with BindCraft
Pačesa, M., Nickel, L., Schellhaas, C., Schmidt, J. H., Pyatova, E., Kissling, L., Barendse, P., Choudhury, J., Kapoor, S., Alcaraz‐Serna, A., et al. (2025) · Nature · doi:10.1038/s41586-025-09429-6 · Cited by 378 · see citing works
Clustering predicted structures at the scale of the known protein universe
Barrio-Hernandez, I., Yeo, J., Jänes, J., Mirdita, M., Gilchrist, C. L. M., Wein, T., Varadi, M., Velankar, S., Beltrao, P., & Steinegger, M. (2023) · Nature · doi:10.1038/s41586-023-06510-w · Cited by 362 · see citing works
Self-citationScalable emulation of protein equilibrium ensembles with generative deep learning
Lewis, S., Hempel, T., Jiménez-Luna, J., Gastegger, M., Xie, Y., Foong, A. Y. K., Satorras, V. G., Abdin, O., Veeling, B. S., Zaporozhets, I., et al. (2025) · Science · doi:10.1126/science.adv9817 · Cited by 335 · see citing works
Empowering biomedical discovery with AI agents
Gao, S., Fang, A., Huang, Y., Giunchiglia, V., Noori, A., Schwarz, J. R., Ektefaie, Y., Kondic, J., & Zitnik, M. (2024) · Cell · doi:10.1016/j.cell.2024.09.022 · Cited by 315 · see citing works
Bilingual language model for protein sequence and structure
Heinzinger, M., Weissenow, K., Sanchez, J. G., Henkel, A., Mirdita, M., Steinegger, M., & Rost, B. (2024) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqae150 · Cited by 311 · see citing works
Self-citationAlphaFold Protein Structure Database and 3D-Beacons: New Data and Capabilities
Fleming, J., Magaña, P., Nair, S. S., Tsenkov, M. I., Bertoni, D., Pidruchna, I., Afonso, M. Q. L., Midlik, A., Paramval, U., Žídek, A., et al. (2025) · Journal of Molecular Biology · doi:10.1016/j.jmb.2025.168967 · Cited by 311 · see citing works
Self-citationSaProt: Protein Language Modeling with Structure-aware VocabularyPreprint
Su, J., Han, C., Zhou, Y., Shan, J., Zhou, X., & Yuan, F. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.10.01.560349 · Cited by 251 · see citing works
Computational design of serine hydrolases
Lauko, A., Pellock, S. J., Sumida, K. H., Anishchenko, I., Juergens, D., Ahern, W., Jeung, J., Shida, A. F., Hunt, A., Kalvet, I., et al. (2025) · Science · doi:10.1126/science.adu2454 · Cited by 235 · see citing works
Fine-tuning protein language models boosts predictions across diverse tasks
Schmirler, R., Heinzinger, M., & Rost, B. (2024) · Nature Communications · doi:10.1038/s41467-024-51844-2 · Cited by 229 · see citing works
PADLOC: a web server for the identification of antiviral defence systems in microbial genomes
Payne, L. J., Meaden, S., Mestre, M. R., Palmer, C., Toro, N., Fineran, P. C., & Jackson, S. A. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkac400 · Cited by 227 · see citing works
Machine Learning-Guided Protein Engineering
Kouba, P., Kohout, P., Haddadi, F., Bushuiev, A., Samusevich, R., Sedlář, J., Damborský, J., Pluskal, T., Šivic, J., & Mazurenko, S. (2023) · ACS Catalysis · doi:10.1021/acscatal.3c02743 · Cited by 214 · see citing works
RCSB protein Data Bank: exploring protein 3D similarities via comprehensive structural alignments
Bittrich, S., Segura, J., Duarte, J. M., Burley, S. K., & Rose, Y. (2024) · Bioinformatics · doi:10.1093/bioinformatics/btae370 · Cited by 212 · see citing works
Design of highly functional genome editors by modelling CRISPR–Cas sequences
Ruffolo, J. A., Nayfach, S., Gallagher, J. P., Bhatnagar, A., Beazer, J., Hussain, R., Russ, J., Yip, J., Hill, E., Pačesa, M., et al. (2025) · Nature · doi:10.1038/s41586-025-09298-z · Cited by 212 · see citing works
Evaluation of AlphaFold-Multimer prediction on multi-chain protein complexes
Zhu, W., Shenoy, A., Kundrotas, P., & Elofsson, A. (2023) · Bioinformatics · doi:10.1093/bioinformatics/btad424 · Cited by 202 · see citing works
AlphaFold2 structures guide prospective ligand discovery
Lyu, J., Kapolka, N. J., Gumpper, R. H., Alon, A., Wang, L., Jain, M. K., Barros-Álvarez, X., Sakamoto, K., Kim, Y., DiBerto, J. F., et al. (2024) · Science · doi:10.1126/science.adn6354 · Cited by 182 · see citing works
Language models generalize beyond natural proteins
Verkuil, R., Kabeli, O., Du, Y., Wicky, B. I. M., Milles, L. F., Dauparas, J., Baker, D., Ovchinnikov, S., Sercu, T., & Rives, A. (2022) · doi:10.1101/2022.12.21.521521 · Cited by 181 · see citing works
CASTpFold: Computed Atlas of Surface Topography of the universe of protein Folds
Ye, B., Tian, W., Wang, B., & Liang, J. (2024) · Nucleic Acids Research · doi:10.1093/nar/gkae415 · Cited by 180 · see citing works
Computational design of soluble and functional membrane protein analogues
Goverde, C. A., Pačesa, M., Goldbach, N. M., Dornfeld, L., Balbi, P. E. M., Georgeon, S., Rosset, S., Kapoor, S., Choudhury, J., Dauparas, J., et al. (2024) · Nature · doi:10.1038/s41586-024-07601-y · Cited by 161 · see citing works
Uncovering new families and folds in the natural protein universe
Durairaj, J., Waterhouse, A. M., Mets, T., Brodiazhenko, T., Abdullah, M., Studer, G., Tauriello, G., Akdel, M., Andreeva, A. E., Bateman, A. G., et al. (2023) · Nature · doi:10.1038/s41586-023-06622-3 · Cited by 151 · see citing works
Kryshtafovych, A., Schwede, T., Topf, M., Fidelis, K., & Moult, J. (2023) · Proteins: Structure, Function, and Bioinformatics · doi:10.1002/prot.26617 · Cited by 148 · see citing works
Template-assisted covalent modification underlies activity of covalent molecular glues
Li, Y.-D., Ma, M. W., Hassan, M. M., Hunkeler, M., Teng, M., Puvar, K., Rutter, J. C., Lumpkin, R. J., Sandoval, B., Jin, C. Y., et al. (2024) · Nature Chemical Biology · doi:10.1038/s41589-024-01668-4 · Cited by 147 · see citing works
What's next for AlphaFold and the AI protein-folding revolution
Callaway, E. (2022) · Nature · doi:10.1038/d41586-022-00997-5 · Cited by 146 · see citing works
Protein remote homology detection and structural alignment using deep learning
Hamamsy, T., Morton, J. T., Blackwell, R., Berenberg, D., Carriero, N. J., Gligorijević, V., Strauss, C. E. M., Leman, J. K., Cho, K., & Bonneau, R. (2023) · Nature Biotechnology · doi:10.1038/s41587-023-01917-2 · Cited by 145 · see citing works
An atlas of protein homo-oligomerization across domains of life
Schweke, H., Pacesa, M., Levin, T., Goverde, C. A., Kumar, P., Duhoo, Y., Dornfeld, L. J., Dubreuil, B., Georgeon, S., Ovchinnikov, S., et al. (2024) · Cell · doi:10.1016/j.cell.2024.01.022 · Cited by 144 · see citing works
Santana, D. J., Anku, J. A. E., Zhao, G., Zarnowski, R., Johnson, C. J., Hautau, H., Visser, N. D., Ibrahim, A. S., Andes, D., Nett, J. E., et al. (2023) · Science · doi:10.1126/science.adf8972 · Cited by 142 · see citing works
BioLiP2: an updated structure database for biologically relevant ligand–protein interactions
Zhang, C., Zhang, X., Freddolino, L., & Zhang, Y. (2023) · Nucleic Acids Research · doi:10.1093/nar/gkad630 · Cited by 142 · see citing works
Gaborieau, B., Vaysset, H., Tesson, F., Charachon, I., Dib, N., Bernier, J., Dequidt, T., Georjon, H., Clermont, O., Hersen, P., et al. (2024) · Nature Microbiology · doi:10.1038/s41564-024-01832-5 · Cited by 141 · see citing works
Data splitting to avoid information leakage with DataSAIL
Joeres, R., Blumenthal, D. B., & Kalinina, O. V. (2025) · Nature Communications · doi:10.1038/s41467-025-58606-8 · Cited by 141 · see citing works
Exploring structural diversity across the protein universe with The Encyclopedia of Domains
Lau, A. M., Bordin, N., Kandathil, S. M., Sillitoe, I., Waman, V. P., Wells, J., Orengo, C. A., & Jones, D. T. (2024) · Science · doi:10.1126/science.adq4946 · Cited by 140 · see citing works
AlphaFold2 protein structure prediction: Implications for drug discovery
Borkakoti, N., & Thornton, J. M. (2023) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2022.102526 · Cited by 137 · see citing works
Unifying the known and unknown microbial coding sequence space
Vanni, C., Schechter, M. S., Acinas, S. G., Barberán, A., Buttigieg, P. L., Casamayor, E. O., Delmont, T. O., Duarte, C. M., Eren, A. M., Finn, R. D., et al. (2022) · eLife · doi:10.7554/elife.67667 · Cited by 134 · see citing works
Self-citationThe impact of AlphaFold Protein Structure Database on the fields of life sciences
Váradi, M., & Velankar, S. S. (2022) · PROTEOMICS · doi:10.1002/pmic.202200128 · Cited by 132 · see citing works
A survey of k-mer methods and applications in bioinformatics
Moeckel, C., Mareboina, M., Konnaris, M. A., Chan, C. S. Y., Mouratidis, I., Montgomery, A., Chantzi, N., Pavlopoulos, G. A., & Georgakopoulos-Soares, I. (2024) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2024.05.025 · Cited by 131 · see citing works
“K-mers prove effective in capturing and predicting information about protein folding patterns [153], secondary structure elements [153], and tertiary structures [154].”
BackgroundProtein function prediction as approximate semantic entailment
Kulmanov, M., Guzmán-Vega, F. J., Duek Roggli, P., Lane, L., Arold, S. T., & Hoehndorf, R. (2024) · Nature Machine Intelligence · doi:10.1038/s42256-024-00795-w · Cited by 131 · see citing works
Atom-level enzyme active site scaffolding using RFdiffusion2
Ahern, W., Yim, J., Tischer, D., Salike, S., Woodbury, S. M., Kim, D., Kalvet, I., Kipnis, Y., Coventry, B., Altae-Tran, H., et al. (2025) · Nature Methods · doi:10.1038/s41592-025-02975-x · Cited by 131 · see citing works
Site-saturation mutagenesis of 500 human protein domains
Beltran, A., Jiang, X., Shen, Y., & Lehner, B. (2025) · Nature · doi:10.1038/s41586-024-08370-4 · Cited by 126 · see citing works
Tesson, F., Planel, R., Egorov, A. A., Georjon, H., Vaysset, H., Brancotte, B., Néron, B., Mordret, E., Atkinson, G. C., Bernheim, A., et al. (2024) · Peer Community Journal · doi:10.24072/pcjournal.470 · Cited by 120 · see citing works
Multiple protein structure alignment at scale with FoldMason
Gilchrist, C. L. M., Mirdita, M., & Steinegger, M. (2026) · Science · doi:10.1126/science.ads6733 · Cited by 117 · see citing works
Self-citationBirth of protein folds and functions in the virome
Nomburg, J., Doherty, E. E., Price, N., Rabelo, D. B., Zhu, Y. K., & Doudna, J. A. (2024) · Nature · doi:10.1038/s41586-024-07809-y · Cited by 117 · see citing works
Functional and evolutionary significance of unknown genes from uncultivated taxa
del Río, Á. R., Giner‐Lamia, J., Cantalapiedra, C. P., Botas, J., Deng, Z., Hernández-Plaza, A., Palmer, M. M., Santamaría‐Hernando, S., Rodríguez‐Herva, J. J., Ruscheweyh, H., et al. (2023) · Nature · doi:10.1038/s41586-023-06955-z · Cited by 110 · see citing works
McCafferty, C. L., Klumpe, S., Amaro, R. E., Kukulski, W., Collinson, L., & Engel, B. D. (2024) · Cell · doi:10.1016/j.cell.2024.01.005 · Cited by 109 · see citing works
Wang, Y., Gallagher, L. A., Andrade, P. A., Liu, A., Humphreys, I. R., Turkarslan, S., Cutler, K. J., Arrieta-Ortiz, M. L., Li, Y., Radey, M. C., et al. (2023) · Cell · doi:10.1016/j.cell.2023.08.017 · Cited by 108 · see citing works
Mirusviruses link herpesviruses to giant viruses
Gaïa, M., Meng, L., Pelletier, E., Forterre, P., Vanni, C., Fernandez-Guerra, A., Jaillon, O., Wincker, P., Ogata, H., Krupovic, M., et al. (2023) · Nature · doi:10.1038/s41586-023-05962-4 · Cited by 108 · see citing works
Computational Approaches to Predict Protein–Protein Interactions in Crowded Cellular Environments
Grassmann, G., Miotto, M., Desantis, F., Rienzo, L. D., Tartaglia, G. G., Pastore, A., Ruocco, G., Monti, M. M., & Milanetti, E. (2024) · Chemical Reviews · doi:10.1021/acs.chemrev.3c00550 · Cited by 107 · see citing works
Discovery of FoTO1 and Taxol genes enables biosynthesis of baccatin III
McClune, C. J., Liu, C., Wick, C., Peña, R. D. L., Lange, B. M., Fordyce, P. M., & Sattely, E. S. (2025) · Nature · doi:10.1038/s41586-025-09090-z · Cited by 106 · see citing works
Moi, D., Bernard, C., Steinegger, M., Nevers, Y., Langleib, M., & Dessimoz, C. (2025) · Nature Structural & Molecular Biology · doi:10.1038/s41594-025-01649-8 · Cited by 109 · see citing works
Self-citationAlphaFold2 reveals commonalities and novelties in protein structure space for 21 model organisms
Bordin, N., Sillitoe, I., Nallapareddy, V., Rauer, C., Lam, S. D., Waman, V. P., Sen, N., Heinzinger, M., Littmann, M., Kim, S., et al. (2023) · Communications Biology · doi:10.1038/s42003-023-04488-9 · Cited by 103 · see citing works
Self-citationSequencing-based analysis of microbiomes
Pinto, Y., & Bhatt, A. S. (2024) · Nature Reviews Genetics · doi:10.1038/s41576-024-00746-6 · Cited by 101 · see citing works
Genomic language model predicts protein co-regulation and function
Hwang, Y., Cornman, A. L., Kellogg, E. H., Ovchinnikov, S., & Girguis, P. R. (2024) · Nature Communications · doi:10.1038/s41467-024-46947-9 · Cited by 99 · see citing works
The growing repertoire of phage anti-defence systems
Murtazalieva, K., Mu, A., Petrovskaya, A. V., & Finn, R. D. (2024) · Trends in Microbiology · doi:10.1016/j.tim.2024.05.005 · Cited by 99 · see citing works
PLMSearch: Protein language model powers accurate and fast sequence search for remote homology
Liu, W., Wang, Z., You, R., Xie, C., Wei, H., Xiong, Y., Yang, J., & Zhu, S. (2024) · Nature Communications · doi:10.1038/s41467-024-46808-5 · Cited by 96 · see citing works
Albanese, K. I., Barbe, S., Tagami, S., Woolfson, D. N., & Schiex, T. (2025) · Nature Reviews Methods Primers · doi:10.1038/s43586-025-00383-1 · Cited by 97 · see citing works
An all-atom protein generative model
Chu, A. E., Kim, J., Cheng, L., Nesr, G. E., Xu, M., Shuai, R. W., & Huang, P. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2311500121 · Cited by 95 · see citing works
Koh, H. Y., Zheng, Y., Yang, M., Arora, R., Webb, G. I., Pan, S., Li, L., & Church, G. M. (2025) · Nature Reviews Bioengineering · doi:10.1038/s44222-025-00349-8 · Cited by 96 · see citing works
Evolutionary genomics of the emergence of brown algae as key components of coastal ecosystems
Denœud, F., Godfroy, O., Cruaud, C., Heesch, S., Nehr, Z., Tadrent, N., Couloux, A., Brillet-Guéguen, L., Delage, L., Mckeown, D., et al. (2024) · Cell · doi:10.1016/j.cell.2024.10.049 · Cited by 94 · see citing works
Logan: Planetary-Scale Genome Assembly Surveys Life’s DiversityPreprint
Chikhi, R., Lemane, T., Loll‐Krippleber, R., Montoliu-Nerin, M., Raffestin, B., Camargo, A. P., Miller, C. J., Fiamenghi, M. B., Agustinho, D. P., Majidian, S., et al. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.07.30.605881 · Cited by 92 · see citing works
Self-citationLarge language models improve annotation of prokaryotic viral proteins
Flamholz, Z. N., Biller, S. J., & Kelly, L. (2024) · Nature Microbiology · doi:10.1038/s41564-023-01584-8 · Cited by 92 · see citing works
Generative artificial intelligence for de novo protein design
Winnifrith, A., Outeiral, C., & Hie, B. (2024) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2024.102794 · Cited by 91 · see citing works
Design of highly functional genome editors by modeling the universe of CRISPR-Cas sequences
Ruffolo, J. A., Nayfach, S., Gallagher, J., Bhatnagar, A., Beazer, J., Hussain, R., Russ, J., Yip, J., Hill, E., Pacesa, M., et al. (2024) · doi:10.1101/2024.04.22.590591 · Cited by 90 · see citing works
Extracellular cytochrome nanowires appear to be ubiquitous in prokaryotes
Baquero, D. P., Cvirkaitė‐Krupovič, V., Hu, S. S., Fields, J. L., Liu, X., Rensing, C., Egelman, E. H., Krupovìč, M., & Wang, F. (2023) · Cell · doi:10.1016/j.cell.2023.05.012 · Cited by 90 · see citing works
Diverse anti-defence systems are encoded in the leading region of plasmids
Samuel, B., Mittelman, K., Croitoru, S. Y., Haim, M. B., & Burstein, D. (2024) · Nature · doi:10.1038/s41586-024-07994-w · Cited by 90 · see citing works
Enhancing efficiency of protein language models with minimal wet-lab data through few-shot learning
Zhou, Z., Zhang, L., Yu, Y., Wu, B., Li, M., Hong, L., & Tan, P. (2024) · Nature Communications · doi:10.1038/s41467-024-49798-6 · Cited by 89 · see citing works
Predicting protein-protein interactions in the human proteome
Zhang, J., Humphreys, I. R., Pei, J., Kim, J., Choi, C., Yuan, R., Durham, J., Liu, S., Choi, H.-J., Baek, M., et al. (2025) · Science · doi:10.1126/science.adt1630 · Cited by 87 · see citing works
A single NLR gene confers resistance to leaf and stripe rust in wheat
Sharma, D., Avni, R., Gutierrez-Gonzalez, J. J., Kumar, R., Sela, H., Prusty, M. R., Shatil‐Cohen, A., Molnár, I., Holušová, K., Said, M., et al. (2024) · Nature Communications · doi:10.1038/s41467-024-54068-6 · Cited by 87 · see citing works
BFVD—a large repository of predicted viral protein structures
Kim, R. S., Levy Karin, E., Mirdita, M., Chikhi, R., & Steinegger, M. (2025) · Nucleic Acids Research · doi:10.1093/nar/gkae1119 · Cited by 86 · see citing works
Self-citationEnhancing alphafold-multimer-based protein complex structure prediction with MULTICOM in CASP15
Liu, J., Guo, Z., Wu, T., Roy, R. S., Quadir, F., Chen, C., & Cheng, J. (2023) · Communications Biology · doi:10.1038/s42003-023-05525-3 · Cited by 86 · see citing works
Song, Y., Yuan, Q., Chen, S., Zeng, Y., Zhao, H., & Yang, Y. (2024) · Nature Communications · doi:10.1038/s41467-024-52533-w · Cited by 84 · see citing works
Bittrich, S., Bhikadiya, C., Bi, C., Chao, H., Duarte, J. M., Dutta, S., Fayazi, M., Henry, J., Khokhriakov, I., Lowe, R., et al. (2023) · Journal of Molecular Biology · doi:10.1016/j.jmb.2023.167994 · Cited by 83 · see citing works
Herrera, L. P. T., Andreassen, S. N., Caroli, J., Rodríguez‐Espigares, I., Kermani, A. A., Keserű, G. M., Kooistra, A. J., Pándy‐Szekeres, G., & Gloriam, D. E. I. (2024) · Nucleic Acids Research · doi:10.1093/nar/gkae1065 · Cited by 84 · see citing works
PLINDER: The protein-ligand interactions dataset and evaluation resourcePreprint
Durairaj, J., Adeshina, Y. O., Cao, Z., Zhang, X., Oleinikovas, V., Duignan, T. J., McClure, Z. D., Robin, X., Studer, G., Kovtun, D., et al. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.07.17.603955 · Cited by 82 · see citing works
AlphaFold two years on: Validation and impact
Kovalevskiy, O. V., Mateos-Garcia, J. C., & Tunyasuvunakool, K. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2315002121 · Cited by 81 · see citing works
Structural diversity of axonemes across mammalian motile cilia
Leung, M. R., Sun, C., Zeng, J., Anderson, J. R., Niu, Q., Huang, W., Noteborn, W. E. M., Brown, A., Zeev‐Ben‐Mordehai, T., & Zhang, R. (2025) · Nature · doi:10.1038/s41586-024-08337-5 · Cited by 80 · see citing works
Scalable protein design using optimization in a relaxed sequence space
Frank, C., Khoshouei, A., Fuβ, L., Schiwietz, D., Putz, D., Weber, L., Zhao, Z., Hattori, M., Feng, S., de Stigter, Y., et al. (2024) · Science · doi:10.1126/science.adq1741 · Cited by 79 · see citing works
Revolutionizing CRISPR technology with artificial intelligence
Kim, M., Go, M.-J., Kang, S.-H., Jeong, S.-H., & Lim, K. (2025) · Experimental & Molecular Medicine · doi:10.1038/s12276-025-01462-9 · Cited by 80 · see citing works
Enzymatic polyethylene biorecycling: Confronting challenges and shaping the future
Jin, J., Arciszewski, J., Auclair, K., & Jia, Z. (2023) · Journal of Hazardous Materials · doi:10.1016/j.jhazmat.2023.132449 · Cited by 78 · see citing works
Wang, R., Zhang, A., Sun, S., Yin, G., Wu, X., Ding, Q., Wang, Q., Chen, F., Wang, S., van Dorp, L., et al. (2024) · Nature Communications · doi:10.1038/s41467-023-44351-3 · Cited by 78 · see citing works
Tesson, F., Huiting, E. D., Wei, L., Ren, J., Johnson, M., Planel, R., Cury, J., Feng, Y., Bondy‐Denomy, J., & Bernheim, A. G. (2024) · Nucleic Acids Research · doi:10.1093/nar/gkae1171 · Cited by 78 · see citing works
Cheng, P., Mao, C., Tang, J., Yang, S., Cheng, Y., Wang, W., Gu, Q., Han, W., Chen, H., Li, S., et al. (2024) · Cell Research · doi:10.1038/s41422-024-00989-2 · Cited by 76 · see citing works
Diffusion models in protein structure and docking
Yim, J., Stärk, H., Corso, G., Jing, B., Barzilay, R., & Jaakkola, T. (2024) · Wiley Interdisciplinary Reviews Computational Molecular Science · doi:10.1002/wcms.1711 · Cited by 76 · see citing works
Rapid and sensitive protein complex alignment with Foldseek-Multimer
Kim, W., Mirdita, M., Karin, E. L., Gilchrist, C. L. M., Schweke, H., Söding, J., Levy, E. D., & Steinegger, M. (2025) · Nature Methods · doi:10.1038/s41592-025-02593-7 · Cited by 75 · see citing works
Self-citationViroid-like colonists of human microbiomes
Zheludev, I. N., Edgar, R. C., Lopez-Galiano, M. J., de la Peña, M., Babaian, A., Bhatt, A. S., & Fire, A. Z. (2024) · Cell · doi:10.1016/j.cell.2024.09.033 · Cited by 73 · see citing works
Updated protein domain annotation of the PARP protein family sheds new light on biological function
Suskiewicz, M. J., Munnur, D., Strømland, Ø., Yang, J.-C., Easton, L. E., Chatrin, C., Zhu, K., Baretić, D., Goffinont, S., Schuller, M., et al. (2023) · Nucleic Acids Research · doi:10.1093/nar/gkad514 · Cited by 79 · see citing works
Computational enzyme design by catalytic motif scaffolding
Braun, M., Tripp, A., Chakatok, M., Kaltenbrunner, S., Fischer, C., Stoll, D., Bijelic, A., Elaily, W., Totaro, M. G., Moser, M., et al. (2026) · Nature · doi:10.1038/s41586-025-09747-9 · Cited by 73 · see citing works
Viroid-like colonists of human microbiomes
Zheludev, I. N., Edgar, R. C., Lopez-Galiano, M. J., de la Peña, M., Babaian, A., Bhatt, A. S., & Fire, A. Z. (2024) · doi:10.1101/2024.01.20.576352 · Cited by 73 · see citing works
Novel machine learning approaches revolutionize protein knowledge
Bordin, N., Dallago, C., Heinzinger, M., Kim, S., Littmann, M., Rauer, C., Steinegger, M., Rost, B., & Orengo, C. (2023) · Trends in Biochemical Sciences · doi:10.1016/j.tibs.2022.11.001 · Cited by 72 · see citing works
Self-citationBiophysics-based protein language models for protein engineering
Gelman, S., Johnson, B., Freschlin, C. R., Sharma, A., D'Costa, S., Peters, J., Gitter, A., & Romero, P. A. (2025) · Nature Methods · doi:10.1038/s41592-025-02776-2 · Cited by 70 · see citing works
Protein structure-informed bacteriophage genome annotation with Phold
Bouras, G., Grigson, S. R., Mirdita, M., Heinzinger, M., Papudeshi, B., Mallawaarachchi, V., Green, R., Kim, R. S., Mihalia, V., Psaltis, A. J., et al. (2026) · Nucleic Acids Research · doi:10.1093/nar/gkaf1448 · Cited by 71 · see citing works
Self-citationGPU-accelerated homology search with MMseqs2
(2025) · doi:10.1038/s41592-025-02819-8 · Cited by 71 · see citing works
Self-citationDeep learning for protein structure prediction and design—progress and applications
Jänes, J., & Beltrão, P. (2024) · Molecular Systems Biology · doi:10.1038/s44320-024-00016-x · Cited by 75 · see citing works
- Authors
- van Kempen, M., Kim, S. S., Tumescheit, C., Mirdita, M., Lee, J., Gilchrist, C. L. M., Söding, J., & Steinegger, M.
- Affiliations
- Seoul National University (KR)
- Max Planck Institute for Multidisciplinary Sciences (DE)
- Corresponding author
- Söding, Johannes
- Steinegger, Martin
- ORCID
- Venue
- Nature Biotechnology
- Published
- 2023-05-08
- Volume
- 42
- Issue
- 2
- Pages
- 243-246
- ISSN
- 0733-222X (ISSN-L)
- 1087-0156
- 1546-1696
- PMC
- PMC10869269
- OpenAlex
- W4375858802
- Semantic Scholar
- 655fa18c26267f33084647b30fb0964ec199cd97
- Cited by
- 2,596
- References
- 48
- Citations by year
- 2026: 828 · 2025: 917 · 2024: 592 · 2023: 211 · 2022: 44 · 2012: 2
- FWCI
- 227.22
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- Nature Biotechnology (publishedVersion, cc-by)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- MPG.PuRe (Max Planck Society) (submittedVersion, other-oa)
- GoeScholar The Publication Server of the Georg-August-Universität Göttingen (Georg-August-Universität Göttingen) (submittedVersion, cc-by)
- Seoul National University Open Repository (Seoul National University) (submittedVersion)
- Funders
- National Research Foundation of Korea (2019R1-A6A1-A10073437, 2019R1A6A1A10073437, 2020M3A9G7103933, 2021R1C1C102065, 2021M3A9I4021220)
- German ministry for education and research (BMBF)
- Samsung DS research fund, Creative-Pioneering Researchers Program through Seoul National University
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
- Physical Sciences › Materials Science › Materials Chemistry › Enzyme Structure and Function
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Machine Learning in Bioinformatics
- Keywords
- protein structure search, Foldseek, structural alignment, structural alphabet, protein structure database, computational efficiency
- MeSH terms
- Algorithms, Amino Acids, Proteins, Software, Databases, Protein
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.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.
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
