QSAR modeling: where have you been? Where are you going to?

Quantitative structure-activity relationship modeling is one of the major computational tools employed in medicinal chemistry. However, throughout its entire history it has drawn both praise and criticism concerning its reliability, limitations, successes, and failures. In this paper, we discuss (i) the development and evolution of QSAR; (ii) the current trends, unsolved problems, and pressing challenges; and (iii) several novel and emerging applications of QSAR modeling. Throughout this discussion, we provide guidelines for QSAR development, validation, and application, which are summarized in best practices for building rigorously validated and externally predictive QSAR models. We hope that this Perspective will help communications between computational and experimental chemists toward collaborative development and use of QSAR models. We also believe that the guidelines presented here will help journal editors and reviewers apply more stringent scientific standards to manuscripts reporting new QSAR studies, as well as encourage the use of high quality, validated QSARs for regulatory decision making.

Frozen out: molecular modeling in the age of cryocrystallography.

As molecular modellers we need to remember that the flexibility of a protein is necessary for it to function. Unfortunately, this flexibility is not readily apparent from the seductive molecular graphics rendering of cryocrystallographic results.

A crowdsourcing evaluation of the NIH chemical probes.

Between 2004 and 2008, the US National Institutes of Health Molecular Libraries and Imaging initiative pilot phase funded 10 high-throughput screening centers, resulting in the deposition of 691 assays into PubChem and the nomination of 64 chemical probes. We crowdsourced the Molecular Libraries and Imaging initiative output to 11 experts, who expressed medium or high levels of confidence in 48 of these 64 probes.

Application of belief theory to similarity data fusion for use in analog searching and lead hopping.

A wide variety of computational algorithms have been developed that strive to capture the chemical similarity between two compounds for use in virtual screening and lead discovery. One limitation of such approaches is that, while a returned similarity value reflects the perceived degree of relatedness between any two compounds, there is no direct correlation between this value and the expectation or confidence that any two molecules will in fact be equally active. A lack of a common framework for interpretation of similarity measures also confounds the reliable fusion of information from different algorithms. Here, we present a probabilistic framework for interpreting similarity measures that directly correlates the similarity value to a quantitative expectation that two molecules will in fact be equipotent. The approach is based on extensive benchmarking of 10 different similarity methods (MACCS keys, Daylight fingerprints, maximum common subgraphs, rapid overlay of chemical structures (ROCS) shape similarity, and six connectivity-based fingerprints) against a database of more than 150,000 compounds with activity data against 23 protein targets. Given this unified and probabilistic framework for interpreting chemical similarity, principles derived from decision theory can then be applied to combine the evidence from different similarity measures in such a way that both capitalizes on the strengths of the individual approaches and maintains a quantitative estimate of the likelihood that any two molecules will exhibit similar biological activity.

A bioavailability score.

Responding to a demonstrated need for scientists to forecast the permeability and bioavailability (F) properties of compounds before their purchase, synthesis, or advanced testing, we have developed a score that assigns the probability that a compound will have F > 10% in the rat. Neither the rule-of-five, log P, log D, nor the combination of the number of rotatable bonds and polar surface area successfully categorized compounds. Instead, different properties govern the bioavailability of compounds depending on their predominant charge at biological pH. The fraction of anions with >10% F falls from 85% if the polar surface area (PSA) is < or = 75 A(2), to 56% if 75 < PSA < 150 A(2), to 11% if PSA is > or = 150 A(2). On the other hand, whereas 55% of the neutral, zwitterionic, or cationic compounds that pass the rule-of-five have >10% F, only 17% of those that fail have > 10% F. This same categorization distinguishes compounds that are poorly permeable from those that are permeable in Caco-2 cells. Further validation is provided with human bioavailability values from the literature.

Ligand binding to domain-3 of human serum albumin: a chemometric analysis.

A detailed chemometric analysis of ligand binding to domain-3A of human serum albumin is described. NMR and fluorescence data on a set of 889 chemically diverse compounds were used to develop a group contribution model based on 74 chemical fragments that is in good agreement with the experimental data (R2 = 0.94, Q2 = 0.90). The structural descriptors used in this analysis comprise a convenient look-up table for quantitatively estimating the effect that a particular group will have on albumin binding. This information can be valuable for optimizing a particular series of compounds for drug development.

Do structurally similar molecules have similar biological activity?

To design diverse combinatorial libraries or to select diverse compounds to augment a screening collection, computational chemists frequently reject compounds that are > or =0.85 similar to one already chosen for the combinatorial library or in the screening set. Using Daylight fingerprints, this report shows that for IC(50) values determined as a follow-up to 115 high-throughput screening assays, there is only a 30% chance that a compound that is > or = 0.85 (Tanimoto) similar to an active is itself active. Although this enrichment is greater than that found with random screening and docking to three-dimensional structures, this low fraction of actives within similar compounds occurs not only because of deficiencies in the Daylight fingerprints and Tanimoto similarity calculations but also because similar compounds do not necessarily interact with the target macromolecule in similar ways. The current study emphasizes the statistical or probabilistic nature of library design and that perfect results cannot be expected.