Pharmacolgical and biological annotations enhance functional residues prediction

• Autores: Paolo Maietta
• Directores de la Tesis: Michael Tress (dir. tes.)
• Lectura: En la Universidad Autónoma de Madrid ( España ) en 2017
• Idioma: inglés
• Tribunal Calificador de la Tesis: Federico Gago Badenas (presid.), Gonzalo Gómez López (secret.), Ana María Rojas Mendoza (voc.)
• Materias:
  • Ciencias de la vida
    • Biología molecular
• Enlaces
  • Tesis en acceso abierto en: Biblos-e Archivo
• Resumen

  • The recent exponential growth of Next Generation Sequencing data has led to a noticeable increase of deposited protein sequences without annotated function. As a result of this, a number of computational methods to automatically infer function have been published in the recent years. However, function prediction is complicated by a number of factors. To begin with, the definition of function itself is not straightforward.

    Furthermore it has been demonstrated that the relationship between sequence, structure and function is not linear and this means that automatic inference using global sequence homology and/or structure homology is not easily applicable. In addition, many proteins have multiples roles that depend on different factors.

    Often the most interesting functional information is to be found at the residue level. We have developed tools to predict functional residues in the CNIO and this thesis takes as starting point two of these, FireDB and firestar. FireDB is a database that extracts information about ligand binding sites and catalytic residues directly from the Protein Data Bank (PDB) and Catalytic Site Atlas (CSA). firestar is a tool that takes advantage of this structured data to predict binding sites for proteins of unknown function and/or structure.

    This thesis describes the many improvements applied to these two tools. For the FireDB database, functional and chemical information has been added for all PDB binding compounds. Ligands have been manually annotated for their biological relevance. In addition the biological relevance of every conserved binding site has been analysed automatically, complementing the existing evaluation schema. All these changes have been included in the revised schema of the database.

    The sensitivity of firestar functional residue prediction has been increased by the addition of a new search method. At the same time, specificity has been improved by examining the data generated in the Critical Assessment of protein Structure Prediction (CASP) experiments. The new parameters were tested in the tenth CASP edition and the final results are presented.

    Finally this thesis describes the incorporation of FireDB and firestar into other tools.

    firestar was used in conjunction with a SIAM, a function prediction algorithm based on homology, in the context of the second edition of the Critical Assessment of protein Function Annotation (CAFA) experiment. FireDB and firestar were used for a study of the functional coherence of the Pfam database protein families. Finally the two methods have been integrated along with other computational methods in the APPRIS database and web services, which provide annotations for alternative splice isoforms and identify principal isoforms for protein coding genes