I’ll review the principles and implementation of the Statistical Coupling Analysis (SCA), an approach to infer groups of coevolving amino acids from the multiple sequence alignment of a protein family.

  The last decades witnessed an explosion of large-scale biological datasets whose analyses require the continuous development of innovative algorithms. Many of these high-dimensional datasets are related to large biological networks with few or no experimentally proven interactions. A striking example lies in the recent gut bacterial studies that provided researchers with a plethora of information sources. Despite a deeper knowledge of microbiome composition, inferring bacterial interactions remains a critical step that encounters significant issues, due in particular to high-dimensional settings, unknown gut bacterial taxa and unavoidable noise in sparse datasets. Such data type make any a priori choice of a learning method particularly difficult and urge the need for the development of new scalable approaches. We propose a consensus method based on spectral decomposition, named Spectral Consensus Strategy, to reconstruct large networks from high-dimensional datasets. This novel unsupervised approach can be applied to a broad range of biological networks and the…

Corynebacterium pseudotuberculosis is a facultative intracellular pathogen and the causative agent of several infectious and contagious chronic diseases, including caseous lymphadenitis, ulcerative lymphangitis, mastitis, and edematous skin disease, in a broad spectrum of hosts. Also, C. pseudotuberculosis infectionspose a rising worldwide economic problem in ruminants. The complete genome sequences of 15 C. pseudotuberculosis strains isolated from different hosts and countries were comparatively analyzed in severalworks from our group. In these works, using pan-genomics, genome plasticity and a novel subtractive genomics approach.  We identified: a clonal aspect of the strains that is different in each biovar; 16 pathogenicity islands that present specific patterns of deletion in each biovar, where the pili clusters of genes are only complete on biovar ovis; and, potential conserved common drug and vaccine targets of these strains along with other Corynebacterium, Mycobacterium and Nocardia (CMN) group of human pathogens considering goat, sheep, bovine, horse, and human as the most affected hosts. Two Corynebacterium-specific ( mscL and resB)and one broad-spectrum ( rpmB ) novel targets were also identified. Six principal targets were subjected to virtual screening, and one compound was found to be…

The extraordinary successes achieved in the last years by bioinformatics, structural biology, and functional studies of proteins have produced an unprecedented amount of data that holds promise to shed light on the workings of these fundamental cell’s molecular machines. In particular theoretical and computational approaches produced powerful methods to explore, piecewise, the renowned paradigm sequence-structure-function, rationalizing the microscopic determinants of protein folding and providing interpretations and testable hypothesis for experiments concerning protein dynamics and function. But is it possible to recapitulate the entire flow of information contained in the paradigm relying only on homologous sequences of the protein of interest? To address this question, I present here a new method based on a network analysis of co-evolutionary coupling aimed at identify groups of residues, named Evolutionary Domains (EDs), which evolved concertedly in a protein family. When applied to a curated sequence database of hundreds of protein families, these clusters result spatially separated…

Underdetermined linear systems of equations are ubiquitous in many fields ranging from biology, to information technology. In recent years, different techniques and approximations have been proposed. In this presentation, I will mostly concentrate on Expectation Propagation (EP) as a general tool for dealing with this problem under a most general structure of the priors. I will start by introducing the formalism with 3 concrete applications: -sampling metabolic networks-tomography-inference of brain activity from eeg data I will then outline the analytical derivation of the algorithm. Finally I will discuss some results on metabolic sampling.

Quantifying the relationship between a biomolecule's genetic sequence and its biological function is a fundamental problem that addresses how complex molecules can evolve. Recent technologies have allowed the collection of large numbers of sequence-function pairs. However, exploiting this data for biophysical insight is challenging, as the data is a sparse sampling of the combinatorially large space of possible sequences. With statistical models inferred from these datasets, I show how a small number of intermediate molecular phenotypes can explain many aspects of sequence-function relationship. From a library of mutated promoter sequences and expression measurements in e.coli lac operon I infer detailed physical interactions between two regulatory proteins. Using a different heuristic approach, I infer fold stability from an thousands of mutated variants of green fluorescent protein, and show how enzymatic activity can be quantified in beta-lactamase. Finally, with a thermodynamic model I infer a detailed energy landscape of a small bacterial protein and…

The technological advances underlying the screening of large combinatorial libraries in high-throughput mutational scans deepen our understanding of adaptive protein evolution and boost its applications in protein design. However, many selection signals can be present in one experiment, entangled with the phenotype of interest. I will show how Restricted Boltzmann Machines (RBM), a powerful yet simple generative model from machine learning, can be trained on all the rounds of a sequence screening experiment, capturing an accurate map of the genotype-to-fitness relationship exercised in the experiment. We discuss how to integrate selections with more than one phenotype, and demonstrate that the model can disentangle multiple signals and predict multi-specific variants. The method fits remarkably well several datasets, with robustness to experimental and sampling noise. I will conclude with other applications of RBMs to screening data.