Application of Belief Theory to Similarity Data Fusion for Use in Analog Searching and Lead Hopping

Abstract

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.

Identification of Cys255 in HIF‐1α as a novel site for development of covalent inhibitors of HIF‐1α/ARNT PasB domain protein–protein interaction

“…We compared COMPOUNDS 1-8 to the entire library of 178 compounds using extended connectivity fingerprints (ECFP6) 35 and setting various levels of Tanimoto similarity, a commonly used technique for identifying compounds similar to a reference compound from a collection of compounds. We did not identify compounds related to the hits summarized in Table II at 80% Tanimoto similarity which would have been expected to show similar biological reactivity.…”

Glossary of terms used in computational drug design, part II (IUPAC Recommendations 2015)

“…belief theory A method to combine probability estimate states that given two or more probabilities P i that a particular event is true, the combined probability of the event is given by [17]:…”

“…Note 2: See also Tversky similarity. Note 3: The value of a Tanimoto similarity of a pair of molecules depends on the molecular fingerprint or structural keys used in the calculation [17]. Note 4: The Tanimoto similarity can also be calculated for molecules described by continuous variables.…”

Shaping a Screening File for Maximal Lead Discovery Efficiency and Effectiveness: Elimination of Molecular Redundancy

“…Similar work, using IC50 level data and with a focus on lead-hopping has been reported by Hajduk and coworkers at Abbott. 38 Based on primary HTS data we defined a compound as active at cut-offs of 50% inhibition in the screen (40% for one of the HTS targets). We chose to use primary HTS data rather than IC50 level data in order to: 1) take a more conservative approach to the estimation of the probability of biological activity at a given Tanimoto similarity and 2) avoid the possibility of compounds being deselected as part of a triage process.…”

“…Thus, fourthly and finally, we introduce here for the first time the concept of molecular redundancy in order to answer this question. Our work builds on activity probability methodology introduced by Nilakantan et al, 35 which was further developed by Harper et al 36 and Lipkin et al 37 as well as work relating chemical similarity to biological activity by Hajduk and Martin 38 Our computational method enables us firstly to determine how many compounds in a given area of chemical space will be needed in order to generate at least one hit in HTS (see below) and secondly to select redundant compounds that are not required.…”