Unified framework for modeling multivariate distributions in biological\n sequences

Le résumé fourni par la source

Revealing the functional sites of biological sequences, such as evolutionary\nconserved, structurally interacting or co-evolving protein sites, is a\nfundamental, and yet challenging task. Different frameworks and models were\ndeveloped to approach this challenge, including Position-Specific Scoring\nMatrices, Markov Random Fields, Multivariate Gaussian models and most recently\nAutoencoders. Each of these methods has certain advantages, and while they have\ngenerated a set of insights for better biological predictions, these have been\nrestricted to the corresponding methods and were difficult to translate to the\ncomplementary domains. Here we propose a unified framework for the\nabove-mentioned models, that allows for interpretable transformations between\nthe different methods and naturally incorporates the advantages and insight\ngained individually in the different communities. We show how, by using the\nunified framework, we are able to achieve state-of-the-art performance for\nprotein structure prediction, while enhancing interpretability of the\nprediction process.\n

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.