Section: New Results

Protein 3D structure

Discovering protein conformations by distance geometry

Participant: A. Mucherino

The distance geometry asks whether a simple weighted undirected graph $G$ can be embedded in a Euclidean space having a predefined dimension $K>0$, so that distances between pairs of embedded vertices are the same as the weights on graph edges. One of the most important applications of the distance geometry can be found in biology, where experimental techniques are able to find estimates of certain distances between atom pairs in molecules. Even if the scientific community is used to employ standardized techniques for the solution of this problem, which are essentially based on heuristic searches, we have recently shown that our combinatorial approach to this problem can be in fact employed for solving biological instances of the distance geometry [17] . This work is in collaboration with international people and researchers from the Pasteur Institut in Paris.

Discretization orders for distance geometry

Participant: A. Mucherino

The concept of discretization order is fundamental for the discretization of the distance geometry, i.e. for reducing the search space of a given distance geometry instance to a discrete (and finite) space. A discretization order is an order on the vertices of the graph G representing an instance of the distance geometry that is able to satisfy the discretization assumptions. Recent research was focused on the problem of finding, for a given distance geometry instance, a suitable discretization order that allows for its discretization [32] . The problem is tackled from a purely theoretical point of view in [33] , while a special order for protein backbones was identified in [27] by creating a path on a "pseudo" de Bruijn graph. In [36] , additional requirements are included during the search for a vertex order, in order to identify discretization orders that are also "optimal". In this work, we used Answer Set Programming (ASP) for identifying optimal partial orders that ensure the discretization of distance geometry instances related to proteins. This work is in collaboration with the Dyliss team, as well as international people.

Structure Similarity Detection

Participants: M. Le Boudic-jamin, R. Andonov

The most commonly used among the various measures of alignment similarity are the internal distances root mean squared deviation (RMSDd ) and the coordinate root mean squared deviation (RMSDc ) . In the paper [18] we introduce a novel approach to find similarities between protein structures. Our algorithm is both internal-distances based and Euclidean-coordinates based (i.e., it uses a rigid transformation to optimally superimpose the two structures). Resulting alignments are guaranteed to score well for both RMSDd and RMSDc , while remaining polynomial. We also replace the goal of finding the largest clique by the one of returning several very dense “near-clique” subgraphs. This choice is strongly justified by the observation that distinct solutions to the structural alignment problem that are close to the optimum are all equally viable from the biological perspective, and hence are all equally interesting from the computation standpoint. Our tool is suitable for detecting similar domains when comparing multi-domain proteins, as well to detect structural repetitions within a single protein and between related proteins [12] .

Automatic Classification of Protein Structure

Participants: M. Le Boudic-jamin, R. Andonov

In this paper [15] we propose a new distance measure for comparing two protein structures based on their contact map representations . We show that our novel measure, which we refer to as the maximum contact map overlap (max-CMO) metric, satisfies all properties of a metric on the space of protein representations. Having a metric in that space allows one to avoid pairwise comparisons on the entire database and, thus, to significantly accelerate exploring the protein space compared to no-metric spaces. We show on a gold standard superfamily classification benchmark set of 6759 proteins that our exact k-nearest neighbor (k-NN) scheme classifies up to 224 out of 236 queries correctly and on a larger, extended version of the benchmark with 60 850 additional structures, up to 1361 out of 1369 queries. Our k-NN classification thus provides a promising approach for the automatic classification of protein structures based on flexible contact map overlap alignments.

Detection of structure repeats in proteins

Participant: M. Le Boudic-jamin, R. Andonov

Almost 25% of proteins contain internal repeats, these repeats may have a major role in the protein function. Furthermore some proteins actually are the same substructure repeated many times, these proteins are solenoids. However, very few protein repeats detection programs exist today. In the paper [29] we present a simple and efficient tool for discovering protein repeats. Our tool is based on protein fragment comparison and clique detection. We show that our tool is able to detect different levels of repetitions and to successfully identify protein tiles.