Welcome to CSAR -- A Resource for Docking and Scoring Development

2011 Benchmark Exercise is Now Open

Blinded Data from Pharma

This year's exercise is devoted to docking and scoring for congeneric series against 4 protein targets. Challenge yourself with a real-world exercise! Instructions can be found here.

Computational chemists need reliable experimental data. The Community Structure-Activity Resource (CSAR) provides experimental datasets of crystal structures and binding affinities for diverse protein-ligand complexes. Some datasets will be generated in house at Michigan while others will be collected from the literature or deposited by academic labs, national centers, and the pharmaceutical industry.

We aim to provide the highest quality data for a diverse collection of proteins and small molecule ligands. We need input from the community in developing our target priorities. Ideal targets will have many high-quality crystal structures (apo and 10-20 bound to diverse ligands) and affinity data for ≥25 compounds that range in size, scaffold, and logP. It is best if the ligand set has several congeneric series that span a broad range of affinity, with low nanomolar to mid-micromolar being most desirable. We prefer Kd data over Ki data over IC50 data (no % activity data). We will determine solubility, pKa, logP/logD data for the ligands whenever possible. We have augmented some donated IC50 data by determining Kon/Koff and ITC data.

CSAR is funded by a U01 grant from the National Institute of General Medical Sciences. The original RFA can be found at http://grants.nih.gov/grants/guide/rfa-files/RFA-GM-08-008.html. Press releases about CSAR can be found at:

• http://www.nigms.nih.gov/News/Results/20081009.htm
• http://www.eurekalert.org/pub_releases/2008-10/uom-uth101008.php

Why should my company donate proprietary data to the public domain? Computational techniques are very successful at enriching hit rates when identifying sets of compounds for experimental testing. However, it is not possible to reliably rank nanomolar-level compounds over those with micromolar affinities. By donating data, it outsources the development of better tools. Pharma has the data, but not the time, to develop improved tools. Second, you have nothing to lose because we are asking for “old” data. Abandoned projects have the kind of data we need, and some could be donated without compromising a company’s competitive advantage on current projects. Third, participation in CSAR can provide visibility in the field. In particular, the donated data could be used to conduct a community-wide blind evaluation of docking and scoring methods. Lastly, there may be a possible financial benefit. Data has value, and it might be possible for the company to declare a charitable donation (of course, this requires consultation with the company’s legal and accounting teams). Our first dataset has been contributed by Abbott (urokinase), and we have reached a legal agreement with GSK to obtain data. We are working with scientists at BMS, Vertex, Pfizer, Merck, Genentech, and Eli Lilly to identify possible depositions. For the community to improve our approaches, we need exceptional datasets to train scoring functions and develop new docking algorithms. That is the goal of the CSAR project.