Development of chemogenomic approaches for prediction of protein - ligand interactions

Abstract : This thesis focuses on the development of bioinformatics methods for the prediction of protein-ligand interactions. The approach used throughout this thesis is to share the known information, both on proteins and on ligands to improve the performance of predictions. The first contribution is the development of a statistical learning method for the prediction of protein - ligands interactions within a family, and is illustrated in then case of GPCRs. This method involves the proposal of new kernels for proteins which take into account the overall similarity of GPCRs based on a sequenced-based hierarchy, and the local similarity of the ligand binding sites of GPCRs based on known 3D structures of known members of this family. A dataset was created to assess the ability of this method to correctly predict the known interactions. The second contribution is the development of a similarity measure between two ligands binding sites from two different (and potentially unrelated) proteins represented by clouds of atoms in 3D. This measure requires pockets alignments using rotations and translations, with the aim of finding the best possible alignment by maximizing the gathering of atoms with similar properties in the nearby regions of space. The performance of this method were measured using a first dataset described in the literature and two others that were created for this purpose. Overall, the results show that chemogenomics approaches display better prediction performances than classical approaches by proteins.
Complete list of metadatas

Cited literature [192 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00679718
Contributor : Bibliothèque Mines Paristech <>
Submitted on : Friday, March 16, 2012 - 11:27:56 AM
Last modification on : Tuesday, November 13, 2018 - 10:10:07 AM
Long-term archiving on : Monday, June 18, 2012 - 4:35:42 PM

Identifiers

  • HAL Id : pastel-00679718, version 1

Citation

Brice Hoffmann. Development of chemogenomic approaches for prediction of protein - ligand interactions. Quantitative Methods [q-bio.QM]. École Nationale Supérieure des Mines de Paris, 2011. English. ⟨NNT : 2011ENMP0074⟩. ⟨pastel-00679718⟩

Share

Metrics

Record views

507

Files downloads

543