Gradual in Silico Filtering for Druglike Substances

Abstract

The suitability of decision trees in comparison to support vector machines for the classification of chemical compounds into drugs and nondrugs was investigated. To account for the requirements upon screening virtual compound libraries, schemes for successive filtering steps with gradual increasing computational cost are outlined. The obtained prediction accuracy was similar between decision trees and support vector machine approaches for the applied compound data sets. By using rapidly computable variables such as druglikeness indices, XlogP, and the molar refractivity, at least 39% of the nondrugs can be filtered out, while retaining more than 83% of the actual drugs. Computationally more demanding descriptors such as specific substructure queries and quantum chemically derived variables can be postponed to subsequent classification schemes for the reduced set of compounds, whereby up to 92% of the nondrugs can be sorted out without loosing considerably more drugs. Using all available computed descriptors simultaneously in the first step did not yield significantly better results. Furthermore, the generated decision trees are used to derive guidelines for the design of druglike substances. The numerical margins found at the branching points suggest several criteria that separate drugs from nondrugs: a molecular weight higher than 230, a molar refractivity higher than 40, and the presence of one or more rings as well as one or more functional groups. Also reported are additionally required parameters to compute values for XlogP, SlogP, and the molar refractivity of boron and silicon containing compounds.

Selecting Relevant Descriptors for Classification by Bayesian Estimates: A Comparison with Decision Trees and Support Vector Machines Approaches for Disparate Data Sets

“…A comprehensive list of these descriptors can be found elsewhere. [20] Additional likeliness scores for each of the three CYP isoforms were generated from the respective substrates and nonsubstrates. The underlying approach is identical to the drug-likeliness index as described earlier in full detail.…”

“…For a detailed explanation of the descriptors see the literature. [20] Property CYP3A4 CYP2D6 CYP2C9 [a] Likeliness scores for the respective CYP isoform were computed from the respective substrates and nonsubstrates (see Section 2) [a] Likeliness scores for the respective CYP isoform were computed from the respective substrates and nonsubstrates (see Section 2) It is well known that particularly large molecules are preferentially metabolized by CYP3A4 due to its larger binding pocket compared to the other CYPs. The molecular volume is likewise the primary criterion in Lewis' manually derived classification scheme.…”

The application of in silico drug-likeness predictions in pharmaceutical research

“…In order to develop more reliable prediction models of drug-likeness, a large number of molecular descriptors and numerous machine learning approaches have been employed, such as support vector machine (SVM) [40][41][42][43]67], neural networks (NN) [37,38,40,41,67,68], genetic algorithm (GA) [69][70][71], recursive partitioning (RP) [45,72], etc. Encouragingly, most prediction models for drug-likeness predictions established by machine learning approaches show satisfactory abilities to discriminate between drug-like and non-drug-like molecules.…”

In SilicoADME/Tox Predictions

“…Once identified, each of these discriminating substructures are tested statistically for enriched activity and compared to the privileged substructures reported in the literature. DTs are also used for the classification of chemical compounds into drug and nondrugs [111,324]. In Schneider et al, DT interpretation suggested the main criteria for separating drugs from nondrugs.…”