Are you sure you want to leave this community? Leaving the community will revoke any permissions you have been granted in this community.
SAMPL6 includes challenges based on aqueous host-guest binding data (binding free energies and, optionally, binding enthalpies) for three different host molecules; and on pKa prediction, for a set of fragment-like molecules. The host-guest systems are useful to test simulation methods, force fields, and solvent models, in the context of binding, without posing the setup issues and computational burden of protein simulations, whereas the pKa prediction challenge is useful in its own right and will pave the way for a future physical property challenge. SAMPL6 will also introduce a new challenge component, the “SAMPLing challenge”, in which computational methods will be evaluated on how efficiently their calculations approach well-converged reference results generated by the organizers. Participants will be provided with machine readable setup files for the molecular systems, including force field setups, along with recommended cutoffs and treatments of long-ranged interactions. The SAMPLing challenge is expected to include one or more cases from each challenge component (host-guest binding on each system; pKa prediction).
As of September 26, all of the information and data files needed to start on the host-guest, SAMPLing, and pKa prediction challenges, are posted, so this challenge is open! Files are hosted at github.com/mobleylab/SAMPL6. Status updates will be posted here and announced by email to the D3R SAMPL list (sign up by joining the challenge here!) and on the D3R Twitter account; we also encourage participants to “watch” the GitHub repository for notifications of file changes/availability and relevant discussions.
Further information on the SAMPL6 challenge components follows. Thanks to Drs. Bruce Gibb (Tulane U.) and Lyle Isaacs (U. Maryland) for providing the host-guest data, and Dr. John Chodera, Mehtap Isik, and Merck for the distribution coefficient data.
The SAMPL6 workshop will be jointly run with D3R, and will take place Feb. 22-23, 2018, in La Jolla, CA.
Gibb Deep Cavity Cavitand (Octa Acids) binding of guests
One host-guest series is based on the Gibb Deep Cavity Cavitands (GDCCs), or octa-acids, previously used in SAMPL4 and SAMPL5. The two hosts, OA and TEMOA (previously OAH and OAME) are identical, except that TEMOA has four additional methyl groups, which alter the shape and depth of the hydrophobic cavity. Both were developed in the laboratory of Dr. Bruce Gibb (Tulane U), who will provide binding free energies and enthalpies, measured by ITC, for eight guest molecules interacting with each host. The measurements are done in 10 mM sodium phosphate buffer at pH 11.7 ± 0.1, and T = 298 K. Host OA is described here: doi:10.1021/ja200633d; and host TEMOA is described here doi:10.1007/s10822-013-9690-2. There are also a number of papers from SAMPL4 and SAMPL5 which discuss calculations for these systems, as summarized, respectively, in doi:10.1007/s10822-014-9735-1 and doi:10.1007/s10822-016-9974-4. Existing benchmark datasets based on these hosts also may be of interest for those preparing to tackle these new complexes: https://github.com/MobleyLab/benchmarksets; this “perpetual” review paper also provides a good introduction to the sampling and experimental issues which are known to be relevant in these systems.
Cucubituril (CB8) binding of guests
This host-guest series is based on the host cucurbituril (CB8), which was used in SAMPL3, as previously summarized (DOI 10.1007/s10822-012-9554-1). CB8 is the eight-membered relative of cucurbit[7uril, which was used in several other prior SAMPL challenges. Data will be provided for ~14 guests, including several FDA approved drugs. Background information on CB8 may be found in a number of publications, including DOI 10.1021/jp2110067, 10.1002/chem.201403405, and 10.1021/ja055013x.
This challenge consists of predicting microscopic and macroscopic pKas of 24 small molecules. These fragment-like small molecules are selected for their similarity to kinase inhibitors and for experimental tractability. Our aim is to evaluate how well current pKa prediction methods perform with drug fragment-like molecules.
In molecules with multiple titratable groups, the protonation state of one group can affect the proton dissociation propensity of another functional group. The microscopic pKa refers to the pKa of deprotonation of a single titratable group while all the other titratable and tautomerizable functional groups of the same molecule are held fixed. The macroscopic pKa defines the acid dissociation constant related to the loss of a proton from a molecule regardless of which functional group the proton is dissociating from, so it doesn't necessarily convey structural information.
pKa measurements were collected using spectrophotometric pKa measurements with a Sirius T3 instrument by Mehtap Isik from the Chodera Lab at Memorial Sloan Kettering with the support of the Merck Rahway Preformulation Department, especially Dorothy Levorse, Timothy Rhodes, and Brad Sherborne.
Complete details on the SAMPL6 pKa challenge are available in the SAMPL6 Github repository, where all information related to input structures and submission directions can be found:
SAMPL6 was originally announced as featuring a logD prediction challenge, but there were difficulties in the collection of experimental data. We were instead able to collect experimental neutral-compound log partition coefficients (logP) for a subset of the SAMPL6 pKa challenge compounds. Thus, these form the basis of SAMPL6 Part II -- a logP prediction challenge commencing immediately. We hope that the logP challenge will be useful for investigating sources of modeling errors that impact solvation, partition, and affinity predictions other than protonation state related errors that were prominent in SAMPL5 logD challenge.
The SAMPL6 logP Challenge consists of predicting the octanol-water partition coefficients of 11 small molecules, which are a subset of the SAMPL6 pKa challenge compounds and resemble fragments of small molecule protein kinase inhibitors. Our aim is to evaluate how well current models can capture the transfer free energy of small molecules between different solvent environments through blind predictions. For physical modeling approaches, this is a means of separating force field accuracy from sampling and protonation state modeling challenges.
Complete details on the SAMPL6 logP Challenge are available in the SAMPL6 GitHub repository, where all information related to input structures and submission directions can be found: