The DOI matches its Crossref record; 66 of the 66 references it cites were found in a registry or an index.
Using deep learning to annotate the protein universe
Open accessPublished version of a preprint
Abstract
Understanding the relationship between amino acid sequence and protein function is a long-standing challenge with far-reaching scientific and translational implications. State-of-the-art alignment-based techniques cannot predict function for one-third of microbial protein sequences, hampering our ability to exploit data from diverse organisms. Here, we train deep learning models to accurately predict functional annotations for unaligned amino acid sequences across rigorous benchmark assessments built from the 17,929 families of the protein families database Pfam. The models infer known patterns of evolutionary substitutions and learn representations that accurately cluster sequences from unseen families. Combining deep models with existing methods significantly improves remote homology detection, suggesting that the deep models learn complementary information. This approach extends the coverage of Pfam by >9.5%, exceeding additions made over the last decade, and predicts function for 360 human reference proteome proteins with no previous Pfam annotation. These results suggest that deep learning models will be a core component of future protein annotation tools.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
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
“These embeddings can capture evolutionary traits of the proteins, as well as functional and structural characteristics.119”
From Advancing Targeted Protein Degradation via Multiomics Profiling and Artificial Intelligence (2023) · cites this paper as [119]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
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
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Gapped BLAST and PSI-BLAST: a new generation of protein database search programs
Altschul, S. (1997) · Nucleic Acids Research · doi:10.1093/nar/25.17.3389 · Cited by 73,302
Distilling the Knowledge in a Neural Network
Hinton, G., Vinyals, O., & Dean, J. (2015) · arXiv · doi:10.48550/arxiv.1503.02531 · Cited by 14,204
Fast and sensitive protein alignment using DIAMOND
Buchfink, B., Xie, C., & Huson, D. H. (2015) · Nature Methods · doi:10.1038/nmeth.3176 · Cited by 14,201
Multi-scale context aggregation by dilated convolutions
Yu, F., & Koltun, V. (2015) · arXiv · doi:10.48550/arxiv.1511.07122 · Cited by 7,738
UniProt: the Universal Protein knowledgebase
Apweiler, R. (2003) · Nucleic Acids Research · doi:10.1093/nar/gkh131 · Cited by 7,593
Research works printed but not linked: 1
- El-Gebali, S., Richardson, L. & Finn, R. Repeats in Pfam. EMBL-EBI Training https://doi.org/10.6019/TOL.Pfam_repeats-t.2018.00001.1 (2018). · DOI
Cited by
Sorted by citations
Paysan‐Lafosse, T., Blum, M., Chuguransky, S., Grego, T., Pinto, B. L., Salazar, G. A., Bileschi, M. L., Bork, P., Bridge, A. J., Colwell, L. J., et al. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkac993 · Cited by 2,457
“A Google Research team is working on the development of ProtENN, a deep-learning method predicting functional annotations for unaligned amino acid sequences using Pfam families as a training set (31).”
Self-citationScientific discovery in the age of artificial intelligence.
Wang, H., Fu, T., Du, Y., Gao, W., Huang, K., Liu, Z., Chandak, P., Liu, S., Van Katwyk, P., Deac, A., et al. (2023) · Nature · doi:10.1038/s41586-023-06221-2 · Cited by 1,823
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 1,240
“ProtENN2, trained on Pfam data, uses convolutional neural networks to annotate protein sequences with Pfam family labels ( 55 ).”
Self-citationMachine-learning-guided directed evolution for protein engineering
Yang, K. K., Wu, Z., & Arnold, F. H. (2019) · Nature Methods · doi:10.1038/s41592-019-0496-6 · Cited by 1,130
Artificial intelligence and illusions of understanding in scientific research
Messeri, L., & Crockett, M. J. (2024) · Nature · doi:10.1038/s41586-024-07146-0 · Cited by 707
“In biology, predictive AI tools are already being used for the automated annotation of protein function 50,51 and cell types 52 .”
Efficient evolution of human antibodies from general protein language models
Hie, B., Shanker, V. R., Xu, D., Bruun, T. U. J., Weidenbacher, P. A. B., Tang, S., Wu, W. C. H., Pak, J. E., & Kim, P. S. (2023) · Nature Biotechnology · doi:10.1038/s41587-023-01763-2 · Cited by 472
“Our techniques can also be used in conjunction with supervised approaches 9,28,33,34,51–54 , and supervising a model over multiple experimental rounds might ultimately lead to higher fitness.”
MGnify: the microbiome sequence data analysis resource in 2023
Richardson, L., Allen, B., Baldi, G., Beracochea, M., Bileschi, M. L., Burdett, T., Burgin, J., Caballero-Pérez, J., Cochrane, G., Colwell, L. J., et al. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkac1080 · Cited by 501
“, (24)), which uses convolutional neural networks to annotate each protein residue in the database with a Pfam family (or clan) label that is then converted into domain calls.”
BackgroundSelf-citationExpanding functional protein sequence spaces using generative adversarial networks
Repecka, D., Jauniskis, V., Karpus, L., Rembeza, E., Rokaitis, I., Zrimec, J., Poviloniene, S., Laurynenas, A., Viknander, S., Abuajwa, W., et al. (2021) · Nature Machine Intelligence · doi:10.1038/s42256-021-00310-5 · Cited by 388
“Although HMMs are powerful and a de facto golden standard for protein functional annotation 39,40 , when used for sequence generation, the first-order HMM did not recapitulate the sequence conservation present in the training data, as the positional entropy error (m.s.e.) of HMM-generated sequences…”
MethodThe 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
Self-citationThe language of proteins: NLP, machine learning & protein sequences
Ofer, D., Brandes, N., & Linial, M. (2021) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2021.03.022 · Cited by 365
“There have been works using contextualized embedding models (e.g. ELMO) on proteins for supervised-learning tasks (such as GO annotation, subcellular localization or structure prediction), as well as clustering sequences based on the learned representations [15,29,54,34].”
MethodOpportunities and Challenges for Machine Learning-Assisted Enzyme Engineering
Yang, J., Li, F.-Z., & Arnold, F. H. (2024) · ACS Central Science · doi:10.1021/acscentsci.3c01275 · Cited by 268
SaProt: 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 264
“…applications, including clinical disease variant prediction (Frazer et al., 2021), fitness landscape prediction (Dallago et al., 2021; He et al., 2024), protein-protein interaction (Nooren & Thornton, 2003), as well as diverse protein function predictions (Bileschi et al., 2022; Yu et al., 2023).”
BackgroundMachine 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 218
Learning functional properties of proteins with language models
Ünsal, S., Ataş, H., Albayrak, M., Turhan, K., Acar, A. C., & Doğan, T. (2022) · Nature Machine Intelligence · doi:10.1038/s42256-022-00457-9 · Cited by 206
Controllable protein design with language models
Ferruz, N., & Höcker, B. (2022) · Nature Machine Intelligence · doi:10.1038/s42256-022-00499-z · Cited by 209
ProteInfer, deep neural networks for protein functional inference
Sanderson, T., Bileschi, M. L., Belanger, D., & Colwell, L. J. (2023) · eLife · doi:10.7554/elife.80942 · Cited by 197
“We also considered an ensemble of neural networks ( Bileschi et al., 2022 ), where the average of the ensemble elements’ predicted probabilities is used as a confidence score, and a naive control, where the number of proteins annotated with a specific term in the training set plays this role (…”
MethodSelf-citationA general model to predict small molecule substrates of enzymes based on machine and deep learning
Kroll, A., Ranjan, S., Engqvist, M. K. M., & Lercher, M. J. (2023) · Nature Communications · doi:10.1038/s41467-023-38347-2 · Cited by 193
“Deep learning models have been used to predict enzyme functions by either predicting their assignment to EC classes 20 – 22 , or by predicting functional domains within the protein sequence 23 .”
MethodUDSMProt: universal deep sequence models for protein classification
Strodthoff, N., Wagner, P., Wenzel, M. A., & Samek, W. (2020) · Bioinformatics · doi:10.1093/bioinformatics/btaa003 · Cited by 192
Transformer-based deep learning for predicting protein properties in the life sciences
Chandra, A. A., Tünnermann, L., Löfstedt, T., & Gratz, R. (2023) · eLife · doi:10.7554/elife.82819 · Cited by 202
“Since recent large-scale DL models have played a crucial role in computational protein property prediction ( Bileschi et al., 2022 ), we describe in this review the most common DL architectures in use today.”
BackgroundECNet is an evolutionary context-integrated deep learning framework for protein engineering
Luo, Y., Jiang, G., Yu, T., Liu, Y., Vo, L. T., Ding, H., Su, Y., Qian, W. W., Zhao, H., & Peng, J. (2021) · Nature Communications · doi:10.1038/s41467-021-25976-8 · Cited by 174
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
“, 2015); or more recently using deep learning approaches (Bileschi et al., 2019; Liu, 2017).”
BackgroundDiffusion models in bioinformatics and computational biology
Guo, Z., Liu, J., Wang, Y., Chen, M., Wang, D., Xu, D., & Cheng, J. (2023) · Nature Reviews Bioengineering · doi:10.1038/s44222-023-00114-9 · Cited by 193
Machine learning for perturbational single-cell omics
Ji, Y., Lotfollahi, M., Wolf, F. A., & Theis, F. J. (2021) · Cell Systems · doi:10.1016/j.cels.2021.05.016 · Cited by 137
Artificial Intelligence in Molecular Medicine
Gomes, B., & Ashley, E. A. (2023) · New England Journal of Medicine · doi:10.1056/nejmra2204787 · Cited by 137
Functional annotation of enzyme-encoding genes using deep learning with transformer layers
Kim, G. B., Kim, J. Y., Lee, J. A., Norsigian, C. J., Palsson, B. Ø., & Lee, S. Y. (2023) · Nature Communications · doi:10.1038/s41467-023-43216-z · Cited by 122
“Therehavebeenotherattemptstocreatehigh-performanceneural networks for biological sequences, such as splitting datasets based on sequence identities 10,39 , using learned protein representations 39 , and incorporating evolutionary information 40,41 .”
Next-Generation TB Vaccines: Progress, Challenges, and Prospects
Li, Z., Ye, Z., Li, L., Yang, L., & Gong, W. (2023) · Vaccines · doi:10.3390/vaccines11081304 · Cited by 126
Antimicrobial resistance crisis: could artificial intelligence be the solution?
Liu, G., Yu, D., Fan, M.-M., Zhang, X., Jin, Z., Tang, C. M., & Liu, X. (2024) · Military Medical Research · doi:10.1186/s40779-024-00510-1 · Cited by 138
“Moreover, when combined with existing methods, DL models can learn complementary information and offer approximately 6.8 million new sequence region annotations, significantly expanding the coverage of Pfam by more than 9.5% [100].”
MethodInfluentialCatPred: a comprehensive framework for deep learning in vitro enzyme kinetic parameters
Boorla, V. S., & Maranas, C. D. (2025) · Nature Communications · doi:10.1038/s41467-025-57215-9 · Cited by 122
“These models offer the promise of high-quality automated functional annotation of sequenced genomes 1 – 3 .”
Convolutions are competitive with transformers for protein sequence pretraining
Yang, K. K., Fusi, N., & Lu, A. X. (2024) · Cell Systems · doi:10.1016/j.cels.2024.01.008 · Cited by 137
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. (2023) · Communications Biology · doi:10.1038/s42003-023-04488-9 · Cited by 103
“…AF2, such as our CATHe predictor, that exploit language models to capture information about the structural contexts of residues, have become widespread and are being applied in the field of protein family classification as they are more powerful and faster than HMM based approaches (24,25,37–39).”
MethodPLMSearch: 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
“More recently, ProtENN 41 uses an ensemble deep learning framework that generated protein sequence embeddings to classify protein domains into Pfam families 42 ; CATHe 43 trains an ANN on embeddings from the PLM ProtT5 38 to detect remote homologs for CATH 44 superfamilies; Embedding-based annotation transfer (EAT) 45…”
MethodAlQuraishi, M., & Sorger, P. K. (2021) · Nature Methods · doi:10.1038/s41592-021-01283-4 · Cited by 96
“We and others have used self-supervised learning to induce representations of protein sequence from very large sequence databases 102-106 , and once learned, applied them to unrelated downstream tasks, including protein design 107 and prediction of protein structure 53,108 and function 109-111 .”
MethodCharacteristics of seismic activity of the Upper Silesian Coal Basin in Poland
Stec, K. (2006) · Geophysical Journal International · doi:10.1111/j.1365-246x.2006.03227.x · Cited by 97
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
Review of AlphaFold 3: Transformative Advances in Drug Design and Therapeutics
Desai, D., Kantliwala, S. V., Vybhavi, J., Ravi, R., Patel, H., & Patel, J. S. (2024) · Cureus · doi:10.7759/cureus.63646 · Cited by 99
Toward an Integrated Machine Learning Model of a Proteomics Experiment
Neely, B. A., Dorfer, V., Martens, L., Bludau, I., Bouwmeester, R., Degroeve, S., Deutsch, E. W., Gessulat, S., Käll, L., Palczynski, P., et al. (2023) · Journal of Proteome Research · doi:10.1021/acs.jproteome.2c00711 · Cited by 79
FuncPEP: A Database of Functional Peptides Encoded by Non-Coding RNAs
Dragomir, M. P., Manyam, G. C., Ott, L. F., Berland, L., Knutsen, E., Ivan, C., Lipovich, L., Broom, B. M., & Călin, G. A. (2020) · Non-Coding RNA · doi:10.3390/ncrna6040041 · Cited by 73
“Comparison of ncPEPs with Known Peptides Encoded by Messenger RNAs of Coding Genes We compared the composition of human ncPEPs with that of human peptides from coding Several criteria for classifying human peptides from a biochemical perspective were retrieved from [26,27]. regions.”
BackgroundMethodInfluentialGotEnzymes: an extensive database of enzyme parameter predictions
Li, F., Chen, Y., Anton, M., & Nielsen, J. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkac831 · Cited by 65
“In addition, we will implement annotations from other databases such as MetaCyc (21) and deep learning-based annotation tools (11,22) to enlarge the coverage of the enzyme-compound pairs, which were generated based only on KEGG database in the initial release.”
Theoretical and Data-Driven Approaches for Biomolecular Condensates
Saar, K. L., Qian, D., Good, L. L., Morgunov, A. S., Collepardo‐Guevara, R., Best, R. B., & Knowles, T. P. J. (2023) · Chemical Reviews · doi:10.1021/acs.chemrev.2c00586 · Cited by 66
Boggust, A., Carter, B., & Satyanarayan, A. (2022) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3490099.3511122 · Cited by 75
“computational biology [6, 7], multimodal learning [56], and computational art [15, 19], evaluating embeddings has informed training procedures and revealed the impact of different training datasets, model architectures, hyperparameters, and model initializations.”
BackgroundSelf-citationMadani, M., Lin, K., & Tarakanova, A. (2021) · International Journal of Molecular Sciences · doi:10.3390/ijms222413555 · Cited by 57
“These 85 pre-extracted biological features were concatenated to the feature maps derived from the flattening layer before fully connected layers in DSResSol (2).”
BackgroundMethodInfluentialEvaluating generalizability of artificial intelligence models for molecular datasets
Ektefaie, Y., Shen, A. N., Bykova, D., Marin, M. G., Žitnik, M., & Farhat, M. R. (2024) · Nature Machine Intelligence · doi:10.1038/s42256-024-00931-6 · Cited by 69
“However, model evaluation is challenging because (1) available molecular sequencing data often capture a small fraction of all possible sequences 4,17 , and (2) sequences evolve and acquire new mutations that are not present in existing datasets.”
A roadmap for the functional annotation of protein families: a community perspective
de Crécy‐Lagard, V., Amorín, R., Arighi, C. N., Babor, J. T., Bateman, A. G., Blaby, I. K., Blaby, C., Bridge, A. J., Burley, S. K., Cleveland, S., et al. (2022) · Database · doi:10.1093/database/baac062 · Cited by 58
“…capacity to rapidly and automatically integrate vast amounts of knowledge, rapidly testing (and correcting) new proposed inferences (81, 82); ML has an incredible capacity for pattern recognition, which includes recognizing inferences that fit these patterns and inferences that do not (42).”
BackgroundSelf-citationApplying machine learning to predict viral assembly for adeno-associated virus capsid libraries
Marques, A. D., Kummer, M., Kondratov, O., Banerjee, A., Moskalenko, O. V., & Zolotukhin, S. (2020) · Molecular Therapy — Methods & Clinical Development · doi:10.1016/j.omtm.2020.11.017 · Cited by 59
Mapping the glycosyltransferase fold landscape using interpretable deep learning
Taujale, R., Zhou, Z., Yeung, W., Moremen, K. W., Li, S., & Kannan, N. (2021) · Nature Communications · doi:10.1038/s41467-021-25975-9 · Cited by 56
“…applied in protein classi fi cation problems such as secondary structure-based HMM searches 26 (Supplementary Table 2), a Long short-term memory (LSTM) model 15 , another CNN-based model 16 , and a more recent transformer-based embeddings model with GDBT classi fi er 17,18 (Supplementary Table 3).”
MethodGenerative power of a protein language model trained on multiple sequence alignments
(2023) · doi:10.7554/elife.79854 · Cited by 43
“Thus, starting from alignment-free methods [24, 70, 71] also constitutes a promising direction.”
Transforming the Language of Life: Transformer Neural Networks for Protein Prediction TasksPreprint
Nambiar, A., Liu, S., Hopkins, M., Heflin, M., Maslov, S., & Ritz, A. (2020) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2020.06.15.153643 · Cited by 54
Příhoda, D., Maritz, J. M., Klempíř, O., Džamba, D., Woelk, C. H., Hazuda, D. J., Bitton, D. A., & Hannigan, G. D. (2020) · Natural Product Reports · doi:10.1039/d0np00055h · Cited by 55
Advancing Targeted Protein Degradation via Multiomics Profiling and Artificial Intelligence
Duran-Frigola, M., Cigler, M., & Winter, G. E. (2023) · Journal of the American Chemical Society · doi:10.1021/jacs.2c11098 · Cited by 50
Responsible AI in biotechnology: balancing discovery, innovation and biosecurity risks
Wheeler, N. E. (2025) · Frontiers in Bioengineering and Biotechnology · doi:10.3389/fbioe.2025.1537471 · Cited by 63
A trimodal protein language model enables advanced protein searches
Su, J., He, Y., You, S., Jiang, S. J., Zhou, X., Zhang, X., Wang, Y., Su, X., Tolstoy, I., Chang, X., et al. (2025) · Nature Biotechnology · doi:10.1038/s41587-025-02836-0 · Cited by 67
“The advent of neural network-based tools 6,8–12 promised a new era, predicting functional labels from predefined vocabularies.”
Are protein language models the new universal key?
Weißenow, K., & Rost, B. (2025) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2025.102997 · Cited by 54
Nallapareddy, V., Bordin, N., Sillitoe, I., Heinzinger, M., Littmann, M., Waman, V. P., Sen, N., Rost, B., & Orengo, C. (2023) · Bioinformatics · doi:10.1093/bioinformatics/btad029 · Cited by 47
Accurate prediction of protein function using statistics-informed graph networks
Jang, Y. J., Qin, Q.-Q., Huang, S., Peter, A. T. J., Ding, X., & Kornmann, B. (2024) · Nature Communications · doi:10.1038/s41467-024-50955-0 · Cited by 50
Fast, sensitive detection of protein homologs using deep dense retrieval
Hong, L., Hu, Z., Sun, S., Tang, X., Wang, J., Tan, Q., Zheng, L., Wang, S., Xu, S. Y., King, I., et al. (2024) · Nature Biotechnology · doi:10.1038/s41587-024-02353-6 · Cited by 46
Mohapatra, S., An, J., & Gómez‐Bombarelli, R. (2022) · Machine Learning Science and Technology · doi:10.1088/2632-2153/ac545e · Cited by 47
DeePVP: Identification and classification of phage virion proteins using deep learning
Fang, Z., Feng, T., Zhou, H., & Chen, M. (2022) · GigaScience · doi:10.1093/gigascience/giac076 · Cited by 42
“For example, Google Research recently designed a tool named ProtCNN, which used the “one-hot” encoding form to represent a protein domain sequence and used a CNN to classify the domain into one member of the Pfam family [28].”
Learning deep representations of enzyme thermal adaptation
Li, G., Buric, F., Zrimec, J., Viknander, S., Nielsen, J., Zelezniak, A., & Engqvist, M. K. M. (2022) · Protein Science · doi:10.1002/pro.4480 · Cited by 44
“In the present study, we used Residual networks 28 , with the model architecture (Figure S1) similar to those that had been applied to protein functional annotation previously 29 .”
MethodNearest neighbor search on embeddings rapidly identifies distant protein relations
Schütze, K., Heinzinger, M., Steinegger, M., & Rost, B. (2022) · Frontiers in Bioinformatics · doi:10.3389/fbinf.2022.1033775 · Cited by 41
DepoScope: Accurate phage depolymerase annotation and domain delineation using large language models
Concha-Eloko, R., Stock, M., Bernard De Baets, C., Briers, Y., Sanjuán, R., Domingo‐Calap, P., & Boeckaerts, D. (2024) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1011831 · Cited by 41
“A notable example is the work by Bileshi et al. (Bileschi et al. 2022) that uses an ensemble of convolutional neural networks to annotate protein domains, which, combined with existing methods, led to a significant expansion of the Pfam database.”
MethodLeveraging deep learning to improve vaccine design
Hederman, A. P., & Ackerman, M. E. (2023) · Trends in Immunology · doi:10.1016/j.it.2023.03.002 · Cited by 41
Malinverni, D., Zamuner, S., Rebeaud, M. E., Barducci, A., Nillegoda, N. B., & De Los Rios, P. (2023) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2218217120 · Cited by 39
The Landscape and Perspectives of the Human Gut Metaproteomics
Sun, Z.-Z., Ning, Z., & Figeys, D. (2024) · Molecular & Cellular Proteomics · doi:10.1016/j.mcpro.2024.100763 · Cited by 39
Machine Learning-Driven Enzyme Mining: Opportunities, Challenges, and Future Perspectives
Moorhoff, F., Zhang, Y., Qiu, S., Dong, W., Medina-Ortiz, D., Zhao, J. B., & Davari, M. D. (2025) · ACS Catalysis · doi:10.1021/acscatal.5c04814 · Cited by 42
“These architectural advances enable models to learn richer representations of protein function and context (Bileschi et al., 2022).”
Qian, H., Wang, Y., Zhou, X.-B., Gu, T., Wang, H., Lyu, H., Li, Z., Li, X., Zhou, H., Guo, C.-C., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-58521-y · Cited by 35
“However, strategies for discovering enzymeswithimprovedcatalyticpropertiescomparedtoexistingones are rarely developed 15,16 .”
Machine learning in predictive biocatalysis: A comparative review of methods and applications
Tripathi, N., Hérisson, J., & Faulon, J. (2025) · Biotechnology Advances · doi:10.1016/j.biotechadv.2025.108698 · Cited by 37
Ramstein, G. P., & Buckler, E. S. (2022) · Genome Biology · doi:10.1186/s13059-022-02747-2 · Cited by 34
“These annotations may be produced by protein representation learning, using techniques like UniRep [13] or more recent models [67, 68]; they may also be produced by 3D structure prediction, using state-of-the-art models like AlphaFold 2 [69].”
MethodConditional generative modeling for de novo protein design with hierarchical functions
Kucera, T., Togninalli, M., & Meng-Papaxanthos, L. (2022) · Bioinformatics · doi:10.1093/bioinformatics/btac353 · Cited by 38
The 2023 Nucleic Acids Research Database Issue and the online molecular biology database collection
Rigden, D., & Fernández, X. M. (2023) · Nucleic Acids Res. · doi:10.1093/nar/gkac1186 · Cited by 27
“The update from the popular MGnify resource for metagenomics data (52), recognising the scale and continued growth of the database, reports interestingly on the adoption of Deep Learning methods to annotate protein sequences with Pfam families (53).”
MethodKim, D. N., McNaughton, A. D., & Kumar, N. (2024) · Bioengineering · doi:10.3390/bioengineering11020185 · Cited by 35
Molecular insights fast-tracked: AI in biosynthetic pathway research
Liao, L., Xie, M., Zheng, X., Zhou, Z., Deng, Z., & Gao, J. (2025) · Natural Product Reports · doi:10.1039/d4np00003j · Cited by 36
Varadi, M., Bordin, N., Orengo, C., & Velankar, S. (2023) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2023.102543 · Cited by 29
“…multiple new predictors based on embeddings from these protein language models were created, tested and validated in various scenarios and were found to outperform established tools for homology detection (including HMMs) [42,43], disorder prediction [44] and ligand-binding prediction [33,45,46].”
Beyond sequence: Structure-based machine learning
Durairaj, J., de Ridder, D., & van Dijk, A. D. J. (2022) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2022.12.039 · Cited by 32
Zhou, Y., Zhao, J., Wen, J., Wu, Z., Dong, Y., & Chen, Y. (2024) · Advanced Science · doi:10.1002/advs.202406912 · Cited by 30
Dou, Z., Sun, Y., Jiang, X., Wu, X., Li, Y., Gong, B., & Wang, L. (2023) · Acta Biochimica et Biophysica Sinica · doi:10.3724/abbs.2023033 · Cited by 30
GOProFormer: A Multi-Modal Transformer Method for Gene Ontology Protein Function Prediction
Kabir, A., & Shehu, A. (2022) · Biomolecules · doi:10.3390/biom12111709 · Cited by 29
Ahmed, B., Haque, M. A., Iquebal, M. A., Jaiswal, S., Angadi, U. B., Kumar, D., & Rai, A. K. (2023) · Frontiers in Plant Science · doi:10.3389/fpls.2022.1008756 · Cited by 29
“, 2020) are being used in building DL for the classification and prediction (Li et al., 2018; Armenteros et al., 2019; Bileschi et al., 2022).”
MethodArtificial intelligence-guided design of lipid nanoparticles for mRNA delivery
Su, K., Qiu, J., Xu, T., & Liu, S. (2025) · Acta Pharmaceutica Sinica B · doi:10.1016/j.apsb.2025.11.029 · Cited by 39
“With the rapid development of computer science, AI stands at the crossroads of multidisciplinary intersections specializing in all kinds of scientific issues related to data-intensity 34 — 37 .”
Ultra-fast protein structure prediction to capture effects of sequence variation in mutation moviesPreprint
Weißenow, K., Heinzinger, M., Steinegger, M., & Rost, B. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.11.14.516473 · Cited by 29
“Distance in embedding space correlates more with protein function than with sequence similarity 15 and can help clustering proteins into families 16-18 .”
Background(2024) · doi:10.1038/s41467-024-52146-3 · Cited by 32
“Likewise, descriptors for larger molecules, such as protein targets, can also be derived, usually gathering features from their amino acid sequences 16 .”
Predicting Antibody Developability from Sequence using Machine LearningPreprint
Chen, X., Dougherty, T., Hong, C., Schibler, R. S., Zhao, Y. C., Sadeghi, R., Matasci, N., Wu, Y.-C., & Kerman, I. (2020) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2020.06.18.159798 · Cited by 26
“Recent work has shown that deep learning models using convolutional neural networks or long short-term memory models can predict protein function from primary sequence (Bileschi et al., 2019; Kulmanov & Hoehndorf, 2020).”
BackgroundDeepSS2GO: protein function prediction from secondary structure
Song, F. V., Su, J., Huang, S., Zhang, N., Li, K., Ni, M., & Liao, M. (2024) · Briefings in Bioinformatics · doi:10.1093/bib/bbae196 · Cited by 26
“Currently, there are multiple protein function classification standards,including Gene Ontology (GO) [4],EC [5],KEGG [6],Pfam [7, 8], etc.”
Heinzinger, M., & Rost, B. (2025) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2025.102986 · Cited by 27
An improved deep learning model for hierarchical classification of protein families
Sandaruwan, P. D., & Wannige, C. T. (2021) · PLoS ONE · doi:10.1371/journal.pone.0258625 · Cited by 25
“In this research work, we use a similar approach to ordered onehot encoding [17,18] for converting an aligned protein sequence to numerical vectors.”
MethodInfluentialProtein language models can capture protein quaternary state
Avraham, O., Tsaban, T., Ben-Aharon, Z., Tsaban, L., & Schueler‐Furman, O. (2023) · BMC Bioinformatics · doi:10.1186/s12859-023-05549-w · Cited by 24
“1] solver = ['sgd', 'adam'] max_iter = [10, 100, 400, 800, 1000, 1500] n_iter_no_change = [5, 10, 15, 20, 30, 40] tol = [0.”
Functional characterization of prokaryotic dark matter: the road so far and what lies ahead
Escudeiro, P., Henry, C. S., & Dias, R. P. M. (2022) · Current Research in Microbial Sciences · doi:10.1016/j.crmicr.2022.100159 · Cited by 24
ProPythia: A Python package for protein classification based on machine and deep learning
Sequeira, A. M., Lousa, D., & Rocha, M. (2021) · Neurocomputing · doi:10.1016/j.neucom.2021.07.102 · Cited by 23
Evaluating large language models for annotating proteins
Vitale, R., Bugnon, L. A., Fenoy, E., Milone, D. H., & Stegmayer, G. S. (2024) · Briefings in Bioinformatics · doi:10.1093/bib/bbae177 · Cited by 22
“In order to assure remote homology (low similarity between training and testing sequences), authors in [5] used single-linkage clustering at 25% identity within each family to build a clustered split of the Pfam-seed data (as suggested in [33]).”
BackgroundMethodResultInfluentialProtein functional site annotation using local structure embeddings
Derry, A., Tartici, A., & Altman, R. B. (2025) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2513219122 · Cited by 21
Possibilities of Using De Novo Design for Generating Diverse Functional Food Enzymes
Wang, X., Xu, K., Tan, Y., Liu, S., & Zhou, J. (2023) · International Journal of Molecular Sciences · doi:10.3390/ijms24043827 · Cited by 22
“Currently, AI-based tools have shown a significant impact on proteomics and structural biology, such as annotating proteins [2], searching functional proteins from the bacterial genome [74], predicting enzyme activity [75], modeling protein structures [22] and designing proteins [32].”
BackgroundNavigating the Expansive Landscapes of Soft Materials: A User Guide for High-Throughput Workflows
Day, E. C., Chittari, S. S., Bogen, M. P., & Knight, A. (2023) · ACS Polymers Au · doi:10.1021/acspolymersau.3c00025 · Cited by 26
Deep embeddings to comprehend and visualize microbiome protein space
Odrzywolek, K., Karwowska, Z., Majta, J., Byrski, A., Milanowska-Zabel, K., & Kosciolek, T. (2022) · Scientific Reports · doi:10.1038/s41598-022-14055-7 · Cited by 22
“Those baseline embeddings are 3-mers with term frequency-inverse document frequency (TFIDF) transformation 51,52 , and amino-acid frequencies vectors (see “Methods”), similarly to seminal works in this field 37,40,45 .”
BackgroundGao, L., Yu, Z., Wang, S., Hou, Y., Zhang, S., Zhou, C.-C., & Wu, X. (2023) · Bioresource Technology · doi:10.1016/j.biortech.2023.129758 · Cited by 21
Investigation of machine learning techniques on proteomics: A comprehensive survey
Sonsare, P. M., & Gunavathi, C. (2019) · Progress in Biophysics and Molecular Biology · doi:10.1016/j.pbiomolbio.2019.09.004 · Cited by 23
Wolf, B., Shehu, P., Brenker, L., von Bachmann, A.-L., Kroell, A.-S. H., Southern, N. T., Holderbach, S., Eigenmann, J., Aschenbrenner, S., Mathony, J., et al. (2025) · Nature Methods · doi:10.1038/s41592-025-02741-z · Cited by 24
“…language models on the other hand, have revolutionized protein engineering by enabling accurate prediction of protein structure 15,16 or functional properties 17–19 based on feature-rich representations and have been successfully applied to various protein engineering and design tasks 20,21 .”
Distinguishing features of fold‐switching proteins
Chakravarty, D., Schafer, J. W., & Porter, L. L. (2023) · Protein Science · doi:10.1002/pro.4596 · Cited by 19
Functional domain annotation by structural similarity
Borujeni, P. M., & Salavati, R. (2024) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqae005 · Cited by 19
The Role of Protein Engineering in Biomedical Applications of Mammalian Synthetic Biology
Bojar, D., & Fussenegger, M. (2019) · Small · doi:10.1002/smll.201903093 · Cited by 20
A large-scale benchmark study of tools for the classification of protein-coding and non-coding RNAs
Singh, D., & Roy, J. K. (2022) · Nucleic Acids Research · doi:10.1093/nar/gkac1092 · Cited by 20
“For instance, ProtTrans (80) is faster than the MMseqs2, ESM (78) achieved performance close that of to hidden Markov model-(HMM) based tools and ProtCNN (82) outperformed the blastp and phmmer tools in a remote homology detection task.”
LambdaPP : Fast and accessible protein‐specific phenotype predictions
(2022) · doi:10.1002/pro.4524 · Cited by 19
“Panel a: residue level features: secondary structure, transmembrane topology, disordered residues, small molecule, nucleic or metal binding residues, residue conservation and average variation (Bernhofer & Rost, 2022; Elnaggar et al., 2021; Ilzhoefer et al., 2022; Littmann, Heinzinger, Dallago, Weissenow, et al., 2021…”
Influential
Metadata
- Authors
- Bileschi, M. L., Belanger, D., Bryant, D. H., Sanderson, T., Carter, B., Sculley, D., Bateman, A., DePristo, M. A., & Colwell, L. J.
- Affiliations
- Google (United States) (US)
- MIT Computer Science and Artificial Intelligence Laboratory (US)
- European Bioinformatics Institute (GB)
- Corresponding author
- Bileschi, Maxwell L.
- Colwell, Lucy J.
- ORCID
- Venue
- Nature Biotechnology
- Published
- 2022-02-21
- Volume
- 40
- Issue
- 6
- Pages
- 932-937
- ISSN
- 0733-222X (ISSN-L)
- 1087-0156
- 1546-1696
- OpenAlex
- W4213112325
- Cited by
- 332
- References
- 59
- Citations by year
- 2026: 39 · 2025: 66 · 2024: 70 · 2023: 73 · 2022: 45
- FWCI
- 20.90
- Type
- journal article
- Language
- en
- Open access
- green
- doi.org
- Locations
- Nature Biotechnology (publishedVersion)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion)
- PubMed (publishedVersion)
- Funders
- Wellcome Trust (210918/Z/18/Z)
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Machine Learning in Bioinformatics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
- Keywords
- protein function prediction, deep learning, Pfam, sequence annotation, remote homology detection, sequence clustering, microbial protein, human proteome, protein families
- MeSH terms
- Deep Learning, Amino Acid Sequence, Humans, Proteome, Databases, Protein, Proteomics, Molecular Sequence Annotation
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.
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
