We’re super excited about the Statistical Assessment of Modeling of Proteins and Ligand (SAMPL) series of blind challenges, and the potential these have for driving real advancements in modeling in our field. These challenges not only provide an opportunity for blind, prospective testing of our methods, but they also provide a crowdsourcing model to drive innovation. Such approaches have a well-documented history of spurring progress in specific areas, such as in the XPrize, the Netflix Prize, or, closer to our field, the DREAM challenges or CASP for protein structure prediction. [More…]
One of the problems we have been very interested in over the years is the problem of ligand binding mode sampling, partly because we find that docking is good at generating plausible ligand binding modes that are stable for relatively long times in MD simulations, but bad at identifying which of these is best. We often then begin MD studies by clustering dock poses, taking members from many different clusters, and then running short MD simulations to identify which are stable. [More…]
Plans are starting to shape up for the SAMPL6 blind prediction challenge, coming later this year! (Our overall SAMPL initiative is still unfunded and we need your input, but we are able to proceed with SAMPL6).
SAMPL6 will focus on host-guest binding on some familiar systems, as well as logD prediction with known (provided) pKa values for a series of small organic compounds. [More…]
For some time, we’ve been involved with running the SAMPL series of blind challenges focused on driving progress in modeling of physical properties and binding (see also the D3R page and our SAMPL NIH proposal) and recently, we’ve sought support from the NIH for this work. However, this is an unconventional type of proposal for the NIH, since the focus is on driving progress of other people’s methods. [More…]