Agnel Praveen Joseph

        The mainstream of my research in Bioinformatics began in 2006 when I joined as a research assistant in Prof.N.Srinivasan's lab of Indian Institute of Science, Bangalore, India. Here, I mainly carried out analysis of specific protein systems largely in collaboration with experimental labs. This involved modeling experiments of human ribosomal proteins, HIV protease & Reverse Transcriptase and plant transcription factors.

       In 2008, I started my thesis work under the guidance of Dr. Alexandre G. de Brevern. Our lab was part of the Genomic and Molecular Bioinformatics team (DSIMB), INSERM. The main focus of my research is the improvement/development of structure comparison and fold prediction methods using Protein Blocks. This work is part of a collaborative project with Pr. N. Srinivasan of Indian Institute of Science (Bangalore, India) and Dr. Bernard Offmann of University of Reunion (Saint-Denis de la Réunion).

       The use of reccurrent local backbone conformations for the comparison of protein structures was explored during my thesis study. The local structure of a protein is usually defined with the help of secondary structural elements like helix, strand and coil. A large majority of about 50% of residues in proteins correspond to the irregular conformation, coil. The coil state can be further analysed in terms of turns,bulges etc. A significant amount of structural variation is also found within the same type of secondary structure, say helices. A more detailed analysis and classification of the local structural elements of proteins can reduce this ambiguity and can present a better picture of local structural states.

       By comparing the phi-psi angle distribution in pentapeptide fragments, a set of 16 frequently occurring local structural motifs in proteins were identified, namely Protein Blocks (PBs) [de Brevern, 2000]. A method for comparing protein structures through the comparison of Protein Blocks , named PBALIGN was developed by Manoj Tyagi. Two protein structures were compared based on the alignment of PB sequences. A dedicated PB susbtitution matrix was developed and a classical Needle-Wunsch dynamic programming algorithm was used to generate PB alignment.

                                                        3D vs 1D(PBs): Flexible Movements


Refined substitution matrices and the use of an anchor-based dynamic programming led to a significant improvement (iPBA). The pairwise alignment method was further extended to develop a multiple structure comparison tool (mulPBA) using a progressive alignment strategy similar to ClustalW. Both the approaches iPBA and mulPBA perform better than other established tools for protein structure comparison. I was also involved in a detailed study of chameleon sequences that represent weak sequence-structure relationships.

        links to chameleon sequence page.

        link to mulPBA.

        iPBA web server

Last modification : 25 September 2011