In silico analyses are increasingly being used to support mode-of-action investigations; however many such approaches do not utilise the large amounts of inactive data held in chemogenomic repositories. The objective this work is concerned with integration bioactivity target prediction orphan compounds produce probability activity and inactivity for a range targets. To end, novel human set was constructed through assimilation over 195 million points deposited ChEMBL PubChem repositories,...

The understanding of the mechanism-of-action (MoA) compounds and prediction potential drug targets play an important role in small-molecule discovery. aim this work was to compare chemical cell morphology information for bioactivity prediction. comparison performed using data from ExCAPE database, image (in form CellProfiler features) Cell Painting set (the largest publicly available images with ∼30,000 compound perturbations), extended connectivity fingerprints (ECFPs) multitask Bayesian...

FAst MEtabolizer (FAME) is a fast and accurate predictor of sites metabolism (SoMs). It based on collection random forest models trained diverse chemical data sets more than 20 000 molecules annotated with their experimentally determined SoMs. Using comprehensive set available data, FAME aims to assess metabolic processes from holistic point view. not limited specific enzyme family or species. Besides global model, dedicated are for human, rat, dog metabolism; prediction phase I II also...

In silico approaches often fail to utilize bioactivity data available for orthologous targets due insufficient evidence highlighting the benefit such an approach. Deeper investigation into orthologue chemical space and its influence toward expanding compound target coverage is necessary improve confidence in this practice.Here we present analysis of ChEMBL PubChem impact on prediction. We highlight number conflicting bioactivities between human orthologues low annotations are overall...

According to Cobanoglu et al., it is now widely acknowledged that the single target paradigm (one protein/target, one disease, drug) has been dominant premise in drug development recent past untenable. More often than not, a drug-like compound (ligand) can be promiscuous - interact with more protein. In years, silico prediction methods promiscuity issue generally approached computationally three main ways: ligand-based methods; target-protein-based and integrative schemes. this study we...

Iterative screening has emerged as a promising approach to increase the efficiency of campaigns compared traditional high throughput approaches. By learning from subset compound library, inferences on what compounds screen next can be made by predictive models, resulting in more efficient screening. One way evaluate is consider cost gain associated with finding an active compound. In this work, we introduce conformal predictor coupled gain-cost function aim maximise iterative Using setup...

In the context of bioactivity prediction, question how to calibrate a score produced by machine learning method into probability binding protein target is not yet satisfactorily addressed. this study, we compared performance three such methods, namely, Platt scaling (PS), isotonic regression (IR), and Venn–ABERS predictors (VA), in calibrating prediction scores obtained from ligand–target comprising Naïve Bayes, support vector machines, random forest (RF) algorithms. Calibration quality was...

Adverse drug reactions (ADRs) are undesired effects of medicines that can harm patients and a significant source attrition in development. ADRs anticipated by routinely screening drugs against secondary pharmacology protein panels. However, there is still lack quantitative information on the links between these off-target proteins reporting humans. Here, we present systematic analysis associations measured predicted vitro bioactivities adverse events (AEs) humans from two sources data: Side...

Abstract Measurements of protein–ligand interactions have reproducibility limits due to experimental errors. Any model based on such assays will consequentially unavoidable errors influencing their performance which should ideally be factored into modelling and output predictions, as the actual standard deviation measurements (σ) or associated comparability activity values between aggregated heterogenous units (i.e., K i versus IC 50 values) during dataset assimilation. However, are usually...

The nest is a protein motif of three consecutive amino acid residues with dihedral angles 1,2-αR αL (RL nests) or 1,2-αL αR (LR nests). Many nests form depression in which an anion δ-negative acceptor atom bound by hydrogen bonds from the main chain NH groups. We have determined extent and nature this bridging database structures using computer program written for purpose. Acceptor anions are pair 40% RL 20% LR nests. Two thirds bridges between groups at Positions 1 3 (N1N3-bridging)-which...

Even though glucuronidations are the most frequent metabolic reactions of conjugation, both in quantitative and qualitative terms, they have rather seldom been investigated using computational approaches. To fill this gap, we used manually collected MetaQSAR reaction database to generate two models for prediction UGT-mediated metabolism, based on molecular descriptors implementing Random Forest algorithm. The first model predicts occurrence was internally validated with a Matthew correlation...

To improve our ability to extrapolate preclinical toxicity humans, there is a need understand and quantify the concordance of adverse events (AEs) between animal models clinical studies. In present work, we discovered 3011 statistically significant associations AEs caused by drugs reported in PharmaPendium database which 2952 were new toxicities encoded different Medical Dictionary for Regulatory Activities terms across species. find plausible testable candidate off-target drug activities...

Pattern classification methods assign an object to one of several predefined classes/categories based on features extracted from observed attributes the (pattern). When L discriminatory for pattern can be accurately determined, problem presents no difficulty. However, precise identification relevant a algorithm (classifier) able categorize real world patterns without errors is generally infeasible. In this case, often cast as devising classifier that minimizes misclassification rate. One way...

Current in vitro models for hepatotoxicity commonly suffer from low detection rates due to incomplete coverage of bioactivity space. Additionally, vivo exposure measures such as Cmax are used screening and unavailable early on. Here we propose a novel rule-based framework extract interpretable biologically meaningful multiconditional associations prioritize end points understand the associated physicochemical conditions. The data this study were derived 673 compounds 361 ToxCast measurements...

In silico protein target deconvolution is frequently used for mechanism-of-action investigations; however existing protocols usually do not predict compound functional effects, such as activation or inhibition, upon binding to their counterparts. This study hence concerned with including effects in prediction. To this end, we assimilated a bioactivity training set 332 targets, comprising 817,239 active data points unknown effect (binding data) and 20,761,260 inactive compounds, along 226,045...

Adverse events resulting from drug therapy can be a cause of withdrawal, reduced and or restricted clinical use, as well major economic burden for society. To increase the safety new drugs, there is need to better understand mechanisms causing adverse events. One way derive mechanistic hypotheses by linking data on with drugs' biological targets. In this study, we have used mining techniques mutual information statistical approaches find associations between reported collected FDA Event...

In the context of bioactivity prediction, question how to calibrate a score produced by machine learning method into reliable probability binding protein target is not yet satisfactorily addressed. this study, we compared performance three such methods, namely Platt Scaling, Isotonic Regression and Venn-ABERS in calibrating prediction scores for ligand-target comprising Naïve Bayes, Support Vector Machines Random Forest algorithms with data available at AstraZeneca (40 million points...

Abstract Predicting the toxicity of a compound preclinically enables better decision making, thereby reducing development costs and increasing patient safety. It is complex issue, but in vitro assays physico-chemical properties compounds can be used to predict clinical toxicity. Neural networks (NNs) are popular predictive tool due their flexibility ability model non-linearities, they prone overfitting therefore not recommended for small data sets. Furthermore, don’t quantify uncertainty...

Abstract A proteolysis targeting chimera (PROTAC) is a new technology that marks proteins for degradation in highly specific manner. During screening, PROTAC compounds are tested concentration-response (CR) assays to determine their potency, and parameters such as the half-maximal concentration (DC 50 ) estimated from fitted CR curves. These used rank compounds, with lower DC values indicating greater potency. However, data often exhibit bi-phasic poly-phasic relationships, making standard...

In the context of bioactivity prediction, question how to calibrate a score produced by machine learning method into reliable probability binding protein target is not yet satisfactorily addressed. this study, we compared performance three such methods, namely Platt Scaling, Isotonic Regression and Venn-ABERS in calibrating prediction scores for ligand-target comprising Naïve Bayes, Support Vector Machines Random Forest algorithms with data available at AstraZeneca (40 million points...

A proteolysis-targeting chimera (PROTAC) is a new technology that marks proteins for degradation in highly specific manner. During screening, PROTAC compounds are tested concentration-response (CR) assays to determine their potency, and parameters such as the half-maximal concentration (DC50) estimated from fitted CR curves. These used rank compounds, with lower DC50 values indicating greater potency. However, data often exhibit biphasic polyphasic relationships, making standard sigmoidal...

In the context of small molecule property prediction, experimental errors are usually a neglected aspect during model generation. The main caveat to binary classification approaches is that they weight minority cases close threshold boundary equivalently in distinguishing between activity classes. For example, pXC50 value 5.1 or 4.9 treated equally important contributing opposing (e.g., 5), even though error may not afford such discriminatory accuracy. This detrimental practice and therefore...