Les protéines BCL-2 jouent un rôle essentiel dans la décision de vie ou de mort des cellules. Elles contrôlent l'induction de l'apoptose (mort cellulaire programmée) par la voie mitochondriale via des fonctions opposées de régulateurs anti- et pro-apoptotiques. Les protéines contenant un ou plusieurs domaines dits d'homologie à Bcl-2 (BHl- 4) sont systématiquement classées dans cette famille. Grâce à une analyse bioinformatique et phylogénétique, nous avons revisité les différents critères d'inclusion dans le groupe de protéines BCL-2 et proposé une nouvelle classification tenant compte des données structurales et évolutives. Cette nouvelle nomenclature distingue : un premier groupe de protéines homologues (dérivant d'un ancêtre commun), partageant une structure 3D semblable à celle de Bcl-2 et pouvant ne posséder aucun motif BH, et un conglomérat, en pleine expansion, regroupant des protéines sans lien phylogénétique apparent et partageant une courte région de similarité de séquence correspondant au motif BH3. Sur la base de ces résultats, nous avons construit un processus, basé sur des profils HMM, pour identifier les protéines appartenant au groupe de protéines BCL-2. Notre processus automatisé est utilisé pour i) récupérer les séquences nucléotidiques et protéiques mensuellement ii) les annoter et iii) les intégrer dans la base de connaissances BCL2DB (« The BCL-2 Database »). Celle-ci est accessible via une interface Web (http://bcl2db.ibcp.fr) qui permet aux chercheurs d'extraire des données et d'effectuer des analyses de séquence / BCL-2 proteins play an essential role in the decision of life or death of animal cells. They control the induction of apoptosis (programmed cell death) in the mitochondrial pathway via regulators having opposite functions: anti- or pro-apoptotic. Proteins containing one or more Bcl-2 homology domains (BHl-4) are systematically classified in this family. Through bioinformatics and phylogenetic analysis, we revisited the different criteria for protein inclusion in the BCL-2 group and proposed a new classification taking into account structural and evolutionary data. This new nomenclature distinguishes a first group of homologous proteins (derived from a common ancestor), sharing a similar 3D structural fold with Bcl-2 and often (but not necessarily) having one or more BH motifs, and a fast expanding conglomerate of proteins without apparent phylogenetic relationships and sharing only a short region of sequence similarity corresponding to the BH3 motif. Based on these results, we built a process based on profiles HMM to identify proteins belonging to the BCL-2 protein group. Our automated process i) recovers on a monthly basis the nucleotide and protein sequences ii) annotates them and iii) integrates this information into BCL2DB ("The BCL-2 Database"). This resource can be accessed via a web interface (http://bcl2db.ibcp.fr) which allows researchers to extract data and perform sequence analysis

BACKGROUND: A long-standing challenge in the post-genomic era of Bioinformatics is the prediction of protein-protein interactions, and ultimately the prediction of protein functions. The problem is intrinsically harder, when only amino acid sequences are available, but a solution is more universally applicable. So far, the problem of uncovering protein-protein interactions has been addressed in a variety of ways, both experimentally and computationally. MOTIVATION: The central problem is: How can protein complexes with solved threedimensional structure be utilized to identify and classify protein binding sites and how can knowledge be inferred from this classification such that protein interactions can be predicted for proteins without solved structure? The underlying hypothesis is that protein binding sites are often restricted to a small number of residues, which additionally often are well-conserved in order to maintain an interaction. Therefore, the signal-to-noise ratio in binding sites is expected to be higher than in other parts of the surface. This enables binding site detection in unknown proteins, when homology based annotation transfer fails. APPROACH: The problem is addressed by first investigating how geometrical aspects of domain-domain associations can lead to a rigorous structural classification of the multitude of protein interface types. The interface types are explored with respect to two aspects: First, how do interface types with one-sided homology reveal convergently evolved motifs? Second, how can sequential descriptors for local structural features be derived from the interface type classification? Then, the use of sequential representations for binding sites in order to predict protein interactions is investigated. The underlying algorithms are based on machine learning techniques, in particular Hidden Markov Models. RESULTS: This work includes a novel approach to a comprehensive geometrical classification of domain interfaces. Alternative structural domain associations are found for 40% of all family-family interactions. Evaluation of the classification algorithm on a hand-curated set of interfaces yielded a precision of 83% and a recall of 95%. For the first time, a systematic screen of convergently evolved motifs in 102.000 protein-protein interactions with structural information is derived. With respect to this dataset, all cases related to viral mimicry of human interface bindings are identified. Finally, a library of 740 motif descriptors for binding site recognition - encoded as Hidden Markov Models - is generated and cross-validated. Tests for the significance of motifs are provided. The usefulness of descriptors for protein-ligand binding sites is demonstrated for the case of "ATP-binding", where a precision of 89% is achieved, thus outperforming comparable motifs from PROSITE. In particular, a novel descriptor for a P-loop variant has been used to identify ATP-binding sites in 60 protein sequences that have not been annotated before by existing motif databases.

Les gènes Hox codent des facteurs de transcription à homéodomain (HD). Bien que ce dernier reconnaisse des séquences similaires in vitro, les protéines Hox achèvent des fonctions hautement spécifiques in vivo. Des séquences protéiques en dehors de l’HD influencent la spécificité d’action des protéines Hox par le recrutement de cofacteurs, dont le mieux caractérisé est Extradenticle (Exd) chez la drosophile. Des travaux récents au sein de notre équipe ont démontré la contribution fonctionnelle de trois motifs de AbdA, aussi bien dans des fonctions Exd-dépendantes qu’à des fonctions Exd-indépendantes. Mon travail de thèse a porté sur la caractérisation de la contribution des motifs protéiques de AbdA dans la sélection puis dans la régulation des gènes cibles en utilisant une approche combinée ChIPseq/RNAseq, dans un contexte Exd-indépendant. Le code ADN identifié nous a renseigné sur la présence d’inputs transcriptionnels additionnels. Ces derniers correspondant à des facteurs de transcription déjà connus, leur présence dans un complexe protéique avec AbdA a été démontrée par des analyses de spectrométrie de masse. Un second volet de mon travail de thèse a été l’identification de modifications post-traductionnelles pouvant rendre compte d’un mécanisme de régulation de l’activité des protéines Hox. Des analyses prédictives in silico, confirmées par des approches biochimiques et des analyses in vivo ont démontré la SUMOylation de AbdA. Ces résultats préliminaires posent les bases pour des travaux futures qui auront pour objectif d’identifier les résidus d’AbdA SUMOylés et d’élucider le rôle de cette modification dans la régulation de l’activité de la protéine AbdA. / Hox genes encode homeodomain-containing transcription factors (HD). Although the HD binds to similar DNA sequences in vitro, Hox proteins display a high functional specificity in vivo. Protein motifs outside of the HD influence Hox specificity through recruiting additional cofactors, with the best characterized being Extradenticle (Exd in Drosophila). Recent evidence from our group has uncovered the functional contribution of AbdA intrinsic motifs to AbdA Exd-dependent functions as well as AbdA Exd-independent functions. My PhD work has aimed to characterize the contribution of AbdA motifs to target gene selection and regulation using a combined approach of ChIPseq/RNAseq in an Exd-independent context. The DNA code identified provides us with new insights about additional transcriptional inputs from additional DNA-binding proteins lying in the vicinity of AbdA recognition sites. Mass spectrometry analysis establishes the occurrence of these additional DNA binding proteins in a multi-protein complex with AbdA. Deciphering the involvement of post-translational modifications in the regulation of Hox protein activity was another aspect of my PhD work. In silico predictive analysis, followed by biochemical approaches and in vivo assays reveal the potential for SUMOylation of AbdA as a potentially important regulatory component of AbdA activity. These preliminary results set the bases for further work aimed at identifying SUMOylated residues on AbdA and the functional relevance of such post-translational modification on AbdA activity regulation.

BACKGROUND: A long-standing challenge in the post-genomic era of Bioinformatics is the prediction of protein-protein interactions, and ultimately the prediction of protein functions. The problem is intrinsically harder, when only amino acid sequences are available, but a solution is more universally applicable. So far, the problem of uncovering protein-protein interactions has been addressed in a variety of ways, both experimentally and computationally. MOTIVATION: The central problem is: How can protein complexes with solved threedimensional structure be utilized to identify and classify protein binding sites and how can knowledge be inferred from this classification such that protein interactions can be predicted for proteins without solved structure? The underlying hypothesis is that protein binding sites are often restricted to a small number of residues, which additionally often are well-conserved in order to maintain an interaction. Therefore, the signal-to-noise ratio in binding sites is expected to be higher than in other parts of the surface. This enables binding site detection in unknown proteins, when homology based annotation transfer fails. APPROACH: The problem is addressed by first investigating how geometrical aspects of domain-domain associations can lead to a rigorous structural classification of the multitude of protein interface types. The interface types are explored with respect to two aspects: First, how do interface types with one-sided homology reveal convergently evolved motifs? Second, how can sequential descriptors for local structural features be derived from the interface type classification? Then, the use of sequential representations for binding sites in order to predict protein interactions is investigated. The underlying algorithms are based on machine learning techniques, in particular Hidden Markov Models. RESULTS: This work includes a novel approach to a comprehensive geometrical classification of domain interfaces. Alternative structural domain associations are found for 40% of all family-family interactions. Evaluation of the classification algorithm on a hand-curated set of interfaces yielded a precision of 83% and a recall of 95%. For the first time, a systematic screen of convergently evolved motifs in 102.000 protein-protein interactions with structural information is derived. With respect to this dataset, all cases related to viral mimicry of human interface bindings are identified. Finally, a library of 740 motif descriptors for binding site recognition - encoded as Hidden Markov Models - is generated and cross-validated. Tests for the significance of motifs are provided. The usefulness of descriptors for protein-ligand binding sites is demonstrated for the case of "ATP-binding", where a precision of 89% is achieved, thus outperforming comparable motifs from PROSITE. In particular, a novel descriptor for a P-loop variant has been used to identify ATP-binding sites in 60 protein sequences that have not been annotated before by existing motif databases.