Cargando…
Generative power of a protein language model trained on multiple sequence alignments
Computational models starting from large ensembles of evolutionarily related protein sequences capture a representation of protein families and learn constraints associated to protein structure and function. They thus open the possibility for generating novel sequences belonging to protein families....
Autores principales: | , , |
---|---|
Formato: | Online Artículo Texto |
Lenguaje: | English |
Publicado: |
eLife Sciences Publications, Ltd
2023
|
Materias: | |
Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10038667/ https://www.ncbi.nlm.nih.gov/pubmed/36734516 http://dx.doi.org/10.7554/eLife.79854 |
_version_ | 1784912130864054272 |
---|---|
author | Sgarbossa, Damiano Lupo, Umberto Bitbol, Anne-Florence |
author_facet | Sgarbossa, Damiano Lupo, Umberto Bitbol, Anne-Florence |
author_sort | Sgarbossa, Damiano |
collection | PubMed |
description | Computational models starting from large ensembles of evolutionarily related protein sequences capture a representation of protein families and learn constraints associated to protein structure and function. They thus open the possibility for generating novel sequences belonging to protein families. Protein language models trained on multiple sequence alignments, such as MSA Transformer, are highly attractive candidates to this end. We propose and test an iterative method that directly employs the masked language modeling objective to generate sequences using MSA Transformer. We demonstrate that the resulting sequences score as well as natural sequences, for homology, coevolution, and structure-based measures. For large protein families, our synthetic sequences have similar or better properties compared to sequences generated by Potts models, including experimentally validated ones. Moreover, for small protein families, our generation method based on MSA Transformer outperforms Potts models. Our method also more accurately reproduces the higher-order statistics and the distribution of sequences in sequence space of natural data than Potts models. MSA Transformer is thus a strong candidate for protein sequence generation and protein design. |
format | Online Article Text |
id | pubmed-10038667 |
institution | National Center for Biotechnology Information |
language | English |
publishDate | 2023 |
publisher | eLife Sciences Publications, Ltd |
record_format | MEDLINE/PubMed |
spelling | pubmed-100386672023-03-25 Generative power of a protein language model trained on multiple sequence alignments Sgarbossa, Damiano Lupo, Umberto Bitbol, Anne-Florence eLife Computational and Systems Biology Computational models starting from large ensembles of evolutionarily related protein sequences capture a representation of protein families and learn constraints associated to protein structure and function. They thus open the possibility for generating novel sequences belonging to protein families. Protein language models trained on multiple sequence alignments, such as MSA Transformer, are highly attractive candidates to this end. We propose and test an iterative method that directly employs the masked language modeling objective to generate sequences using MSA Transformer. We demonstrate that the resulting sequences score as well as natural sequences, for homology, coevolution, and structure-based measures. For large protein families, our synthetic sequences have similar or better properties compared to sequences generated by Potts models, including experimentally validated ones. Moreover, for small protein families, our generation method based on MSA Transformer outperforms Potts models. Our method also more accurately reproduces the higher-order statistics and the distribution of sequences in sequence space of natural data than Potts models. MSA Transformer is thus a strong candidate for protein sequence generation and protein design. eLife Sciences Publications, Ltd 2023-02-03 /pmc/articles/PMC10038667/ /pubmed/36734516 http://dx.doi.org/10.7554/eLife.79854 Text en © 2023, Sgarbossa et al https://creativecommons.org/licenses/by/4.0/This article is distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/) , which permits unrestricted use and redistribution provided that the original author and source are credited. |
spellingShingle | Computational and Systems Biology Sgarbossa, Damiano Lupo, Umberto Bitbol, Anne-Florence Generative power of a protein language model trained on multiple sequence alignments |
title | Generative power of a protein language model trained on multiple sequence alignments |
title_full | Generative power of a protein language model trained on multiple sequence alignments |
title_fullStr | Generative power of a protein language model trained on multiple sequence alignments |
title_full_unstemmed | Generative power of a protein language model trained on multiple sequence alignments |
title_short | Generative power of a protein language model trained on multiple sequence alignments |
title_sort | generative power of a protein language model trained on multiple sequence alignments |
topic | Computational and Systems Biology |
url | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10038667/ https://www.ncbi.nlm.nih.gov/pubmed/36734516 http://dx.doi.org/10.7554/eLife.79854 |
work_keys_str_mv | AT sgarbossadamiano generativepowerofaproteinlanguagemodeltrainedonmultiplesequencealignments AT lupoumberto generativepowerofaproteinlanguagemodeltrainedonmultiplesequencealignments AT bitbolanneflorence generativepowerofaproteinlanguagemodeltrainedonmultiplesequencealignments |