In silico methods of phenotypic screening are necessary to reduce the time and cost of the experimental in vivo screening of anticancer agents through dozens of millions of natural and synthetic chemical compounds. We used the previously developed PASS (Prediction of Activity Spectra for Substances) algorithm to create and validate the classification SAR models for predicting the cytotoxicity of chemicals against different types of human cell lines using ChEMBL experimental data. A training set from 59,882 structures of compounds was created based on the experimental data (IG50, IC50, and % inhibition values) from ChEMBL. The average accuracy of prediction (AUC) calculated by leave-one-out and a 20-fold cross-validation procedure during the training was 0.930 and 0.927 for 278 cancer cell lines, respectively, and 0.948 and 0.947 for cytotoxicity prediction for 27 normal cell lines, respectively. Using the given SAR models, we developed a freely available web-service for cell-line cytotoxicity profile prediction (CLC-Pred: Cell-Line Cytotoxicity Predictor) based on the following structural formula: http://way2drug.com/Cell-line/.

/pdf/clc-pred-a-freely-available-web-service-for-in-silico-2ze9rbj3yy.pdf

CLC-Pred: A freely available web-service for in silico prediction of human cell line cytotoxicity for drug-like compounds.

Here we report PPB2 as a target prediction tool assigning targets to a query molecule based on ChEMBL data. PPB2 computes ligand similarities using molecular fingerprints encoding composition (MQN), molecular shape and pharmacophores (Xfp), and substructures (ECfp4) and features an unprecedented combination of nearest neighbor (NN) searches and Naive Bayes (NB) machine learning, together with simple NN searches, NB and Deep Neural Network (DNN) machine learning models as further options. Although NN(ECfp4) gives the best results in terms of recall in a 10-fold cross-validation study, combining NN searches with NB machine learning provides superior precision statistics, as well as better results in a case study predicting off-targets of a recently reported TRPV6 calcium channel inhibitor, illustrating the value of this combined approach. PPB2 is available to assess possible off-targets of small molecule drug-like compounds by public access at http://gdb.unibe.ch .

/pdf/polypharmacology-browser-ppb2-target-prediction-combining-j8tqj7pgjf.pdf

Polypharmacology Browser PPB2: Target Prediction Combining Nearest Neighbors with Machine Learning.

We present MuSSeL, a multifingerprint similarity search algorithm, able to predict putative drug targets for a given query small molecule as well as to return a quantitative assessment of its bioactivity in terms of Ki or IC50 values. Predictions are automatically made exploiting a large collection of high quality experimental bioactivity data available from ChEMBL (version 22.1) combining, in a consensus-like approach, predictions resulting from a similarity search performed using 13 different fingerprint definitions. Importantly, the herein proposed algorithm is also effective in detecting and handling activity cliffs. A calibration set including small molecules present in the last updated version of ChEMBL (version 23) was employed to properly tune the algorithm parameters. Three randomly built external sets were instead challenged for model performances. The potential use of MuSSeL was also challenged by a prospective exercise for the prediction of five bioactive compounds taken from articles published in the Journal of Medicinal Chemistry just few months ago. The paper emphasizes the importance of implementing multifingerprint consensus strategies to increase the confidence in prediction of similarity search algorithms and provides a fast and easy-to-run tool for drug target and bioactivity prediction.

A New Approach for Drug Target and Bioactivity Prediction: The Multifingerprint Similarity Search Algorithm (MuSSeL).

We describe the current version of the PASS program for prediction of biological activity spectra of organic compounds based on analysis of structure—activity relationships (SAR) for a training set containing information on more than 1000 000 biologically active organic compounds. The average accuracy of prediction for more than 5 000 types of biological activity exceeds a value of 0.96. To analyze quantitative SAR, the GUSAR program was developed. The advantages of GUSAR were demonstrated in a number of computational experiments. The local versions of the PASS and GUSAR programs, as well as 19 freely available web services were developed. The latter are freely accessible via the Internet at http://way2drug.com/dr. The web services at the Way2Drug portal are used by more than 24 000 researchers working in about 100 countries. Currently, more than 830 000 predictions were made, the most promising compounds were selected for chemical synthesis, and priorities for testing their biological activity were established. The PharmaExpert software was developed to analyze the results of the PASS- and GUSAR-based predictions and to search for chemical compounds with necessary biological activity spectra. Combined use of the PASS, GUSAR, and PharmaExpert programs enables an assessment of the pharmacotherapeutic, adverse, and toxic effects of new compounds based on the systems pharmacology.

Computer-aided prediction of biological activity spectra for organic compounds: the possibilities and limitations

Computational methods for target prediction, based on molecular similarity and network-based approaches, machine learning, docking and others, have evolved as valuable and powerful tools to aid the challenging task of mode of action identification for bioactive small molecules such as drugs and drug-like compounds. Critical to discerning the scope and limitations of a target prediction method is understanding how its performance was evaluated and reported. Ideally, large-scale prospective experiments are conducted to validate the performance of a model; however, this expensive and time-consuming endeavor is often not feasible. Therefore, to estimate the predictive power of a method, statistical validation based on retrospective knowledge is commonly used. There are multiple statistical validation techniques that vary in rigor. In this review we discuss the validation strategies employed, highlighting the usefulness and constraints of the validation schemes and metrics that are employed to measure and describe performance. We address the limitations of measuring only generalized performance, given that the underlying bioactivity and structural data are biased towards certain small-molecule scaffolds and target families, and suggest additional aspects of performance to consider in order to produce more detailed and realistic estimates of predictive power. Finally, we describe the validation strategies that were employed by some of the most thoroughly validated and accessible target prediction methods.

/pdf/validation-strategies-for-target-prediction-methods-4grcn4kg55.pdf

Validation strategies for target prediction methods

Drug repurposing provides a non-laborious and less expensive way for finding new human medicines. Computational assessment of bioactivity profiles shed light on the hidden pharmacological potential of the launched drugs. Currently, several freely available computational tools are available via the Internet, which predict multitarget profiles of drug-like compounds. They are based on chemical similarity assessment (ChemProt, SuperPred, SEA, SwissTargetPrediction and TargetHunter) or machine learning methods (ChemProt and PASS). To compare their performance, this study has created two evaluation sets, consisting of (1) 50 well-known repositioned drugs and (2) 12 drugs recently patented for new indications. In the first set, sensitivity values varied from 0.64 (TarPred) to 1.00 (PASS Online) for the initial indications and from 0.64 (TarPred) to 0.98 (PASS Online) for the repurposed indications. In the second set, sensitivity values varied from 0.08 (SuperPred) to 1.00 (PASS Online) for the initial indications and from 0.00 (SuperPred) to 1.00 (PASS Online) for the repurposed indications. Thus, this analysis demonstrated that the performance of machine learning methods surpassed those of chemical similarity assessments, particularly in the case of novel repurposed indications.

How good are publicly available web services that predict bioactivity profiles for drug repurposing

Modern drug discovery is based on the analysis of information about disease mechanisms, molecular targets, and pharmacological substances, the action of which helps to normalize pathological processes Existing chemical and biomedical databases have recently been supplemented with many specialized computational resources that provide the means to estimate physicochemical properties, biological activity, toxicity, metabolism, and other characteristics of organic molecules in the early stages of drug development Easy access to this information via the Internet expands significantly the opportunities for an investigator and, at the same time, poses the problem of selecting the information and computational resources most appropriate to the tasks at hand This paper provides an overview of biomedical and chemical web resources freely available via the Internet and gives recommendations on their optimal use at various stages in the research and development of safer and more effective medicines

Web Resources for Discovery and Development of New Medicines

Molecular property diagnostic suite (MPDS) is a Galaxy-based open source drug discovery and development platform. MPDS web portals are designed for several diseases, such as tuberculosis, diabetes mellitus, and other metabolic disorders, specifically aimed to evaluate and estimate the drug-likeness of a given molecule. MPDS consists of three modules, namely data libraries, data processing, and data analysis tools which are configured and interconnected to assist drug discovery for specific diseases. The data library module encompasses vast information on chemical space, wherein the MPDS compound library comprises 110.31 million unique molecules generated from public domain databases. Every molecule is assigned with a unique ID and card, which provides complete information for the molecule. Some of the modules in the MPDS are specific to the diseases, while others are non-specific. Importantly, a suitably altered protocol can be effectively generated for another disease-specific MPDS web portal by modifying some of the modules. Thus, the MPDS suite of web portals shows great promise to emerge as disease-specific portals of great value, integrating chemoinformatics, bioinformatics, molecular modelling, and structure- and analogue-based drug discovery approaches.

G N Sastry

Papers

How good are publicly available web services that predict bioactivity profiles for drug repurposing

Web Resources for Discovery and Development of New Medicines

Molecular property diagnostic suite (MPDS): Development of disease-specific open source web portals for drug discovery.