Variační nástroj založený na autoenkodéru pro návrh ancestrálních proteinů
| Title in English | Variational Autoencoder-based tool for design of ancestral proteins |
|---|---|
| Authors | |
| Year of publication | 2024 |
| Type | Software |
| MU Faculty or unit | |
| web | Skripty v jazyce Python a datové sady jsou k dispozici na GitHubu |
| Description | To streamline the search for protein sequences with desired properties, we leverage information from protein families through multiple sequence alignments to identify conserved regions and amino acid variability. We show variational autoencoders which can capture the full variation of sequences in multiple sequence alignments by using nonlinear machine learning modeling. We have developed a tool that learns protein representations and can guide researchers in producing improved protein mutants as well as in phylogenetic reconstruction. The use case study is available at https://doi.org/10.26434/chemrxiv-2023-jcds7-v2 |
| Related projects: |