PPIFocusDb

Protein-protein interactions (PPIs) are essential for most cellular processes. Aberrant PPIs contribute to most disease states and as such, PPIs represent a highly populated class of essentially untouched potential targets for drug discovery. About 650,000 PPIs are expected in the human body and most of them are not known. These numbers illustrate the magnitude of the challenge and the situation gets even more overwhelming as PPIs in other organisms are also critical to assess. Overall, these last 30 years, the most successful approach to inhibit protein-protein interactions has been through the use of antibodies and peptides. An alternative approach would be to use small drug-like chemical compounds, but this remains a formidable challenge, time consuming, costly and plagued with technical hurdles. Yet, some recent success stories, in both, academia and in pharmaceutical companies support the idea that drug-like inhibitors of PPIs can be identified. Several overlapping questions are thus at play: How to investigate these PPIs in a cost-efficient manner ? How to rapidly perform proof-of-concept studies on numerous PPIs ? Is it possible to define a chemical space for PPI inhibitors knowing that it is apparently different from the one involving compounds acting on catalytic sites and membrane receptors ? At present, the most efficient approaches to discover PPI inhibitors are experimental high-throughput screening, fragment-based screening and in silico screening. However, these approaches can be very time consuming, not always efficient on PPIs and can be very expensive while in most cases, they do not thoroughly attempt to rationalize the chemical space of the inhibitors of PPIs. In most cases, the compound collections used to perform these studies have not been tailored to this domain of application because we are lacking knowledge about drug-like PPI inhibitors and are thus libraries developed for catalytic sites and membrane receptors. We believe that one of the solutions to this conundrum (high needs but low cost and lack of knowledge) is to design a focused library enriched in putative PPI inhibitors. This small collection could be used in priority for all experimental and proof-of-concept screening studies prior to the use more expensive approaches. We have been developing the beta version of a software that enriches a compound collection in PPI inhibitors through the use of chemoinformatics and relevant machine learning methods. We have challenged our tool with an experimental HTS study performed on the p53-MDM2 protein-protein interaction. We now would like to optimize our mathematical models and analyze the PPI inhibitor chemical space. We then wish to challenge our computer tool by experimentally screening 4 known PPIs with a focused compound collection generated by our in silico filter and a well designed random compound collection while validating the mode of binding of key compounds via X-ray crystallography. At the end of our project we should be able to provide (a) an in silico PPI inhibitor profiler (a profiling software), (b) a focused compound collection ready to use in a HTS lab, and (c) a description of part of the still essentially unknown PPI inhibitor chemical space. This project involves in silico experts, medicinal chemists, structural biologists, molecular biologists and clinicians. It should facilitate the investigation of PPIs, either for chemical biology projects or for drug discovery initiatives. To the best of our knowledge (we survey extensively the field, see www.vls3d.com, links) there are no PPI profiler software nor any focus PPI libraries distributed or commercial. Both, the in silico tool and the focused compound collection have commercial values, they will however be distributed freely upon request to research groups.