Pattern Identification in Sequence Activity Data

National Institute Of Diabetes And Digestive And Kidney Diseases

Investigators

Linked publications, trials & patents

Abstract

We presented a novel method for protein contact inference using Expectation Reflection estimator that infers the influence of the rest of the protein sequence on the presence or absence of a specific residue at a particular position in the amino-acid sequence. We applied this adaptation to 40974 structures to test the efficacy of our approach. In order to predict symmetric observables such as contact maps, we symmetrized the directed coupling matrix and compare its performance to that of mean field analysis (MF) and pseudo-likelihood maximization(PLM). For all ranges of sequence lengths and MSA sizes, our method performs better. Overall, our method outperforms existing methods for 76% of analyzed proteins. A modification of the same method is how we computed the genome-wide covariation of the SARS-CoV-2 genome, showing that the covariation in the genome is clade dependent, for example. Indeed, the clade determinants appear automatically in the analysis of the entire genome as the positions on the genome that are significantly interacting with other positions. We are continuing this work to try to understand the evolution of the viral genome, as well as the changes that may havve occurred after the introductions of vaccines and antiviral drugs. (publication listed with the report on work on the epidemic).