Section: New Results

New insight from SHAPE probing data

Existing computational methods for structure prediction are typically hindered by their assumption of a single structure, and their assumption of orthogonal signals stemming from different reagents. To overcome these limitations, we contributed an integrative approach combining stochastic sampling and structural clustering [17] (journal version pending). In collaboration with ENS Lyon/Univ. Lyon I and Univ. Paris-Descartes, we used this method to model the structure of the HIV-1 gag open-reading frame [4].

We also addressed the problem of binning sets of NGS reads arising from the simultaneous probing, using the SHAPEmap protocol, of variants produced by a error-prone PCR. We proposed a variant of the Expectation-Maximization algorithm [10] to jointly infer maximum-likelihood origins for reads and mutational profiles for each variant.