High-throughput Screening Data Sets Research Articles

Machine Learning-Driven Data Valuation for Optimizing High-Throughput Screening Pipelines.

In the rapidly evolving field of drug discovery, high-throughput screening (HTS) is essential for identifying bioactive compounds. This study introduces a novel application of data valuation, a concept for evaluating the importance of data points based on their impact, to enhance drug discovery pipelines. Our approach improves active learning for compound library screening, robustly identifies true and false positives in HTS data, and identifies important inactive samples in an imbalanced HTS training, all while accounting for computational efficiency. We demonstrate that importance-based methods enable more effective batch screening, reducing the need for extensive HTS. Machine learning models accurately differentiate true biological activity from assay artifacts, streamlining the drug discovery process. Additionally, importance undersampling aids in HTS data set balancing, improving machine learning performance without omitting crucial inactive samples. These advancements could significantly enhance the efficiency and accuracy of drug development.

Machine learning-assisted high-throughput screening for Anti-MRSA compounds.

Antimicrobial resistance is a major public health threat, and new agents are needed. Computational approaches have been proposed to reduce the cost and time needed for compound screening. A machine learning (ML) model was developed for the in silico screening of low molecular weight molecules. We used the results of a high-throughput Caenorhabditis elegans methicillin-resistant Staphylococcus aureus (MRSA) liquid infection assay to develop ML models for compound prioritization and quality control. The compound prioritization model achieved an AUC of 0.795 with a sensitivity of 81% and a specificity of 70%. When applied to a validation set of 22,768 compounds, the model identified 81% of the active compounds identified by high-throughput screening (HTS) among only 30.6% of the total 22,768 compounds, resulting in a 2.67-fold increase in hit rate. When we retrained the model on all the compounds of the HTS dataset, it further identified 45 discordant molecules classified as non-hits by the HTS, with 42/45 (93%) having known antimicrobial activity. Our ML approach can be used to increase HTS efficiency by reducing the number of compounds that need to be physically screened and identifying potential missed hits, making HTS more accessible and reducing barriers to entry.

Lies and Liabilities: Computational Assessment of High-Throughput Screening Hits to Identify Artifact Compounds.

Hits from high-throughput screening (HTS) of chemical libraries are often false positives due to their interference with assay detection technology. In response, we generated the largest publicly available library of chemical liabilities and developed "Liability Predictor," a free web tool to predict HTS artifacts. More specifically, we generated, curated, and integrated HTS data sets for thiol reactivity, redox activity, and luciferase (firefly and nano) activity and developed and validated quantitative structure-interference relationship (QSIR) models to predict these nuisance behaviors. The resulting models showed 58-78% external balanced accuracy for 256 external compounds per assay. QSIR models developed and validated herein identify nuisance compounds among experimental hits more reliably than do popular PAINS filters. Both the models and the curated data sets were implemented in "Liability Predictor," publicly available at https://liability.mml.unc.edu/. "Liability Predictor" may be used as part of chemical library design or for triaging HTS hits.

SDPS-32 A CLINICALLY COMPATIBLE IN VITRO DRUG-SCREENING PLATFORM IDENTIFIES THERAPEUTIC VULNERABILITIES IN PRIMARY CULTURES OF BRAIN METASTASES

Abstract Approximately 20% of cancer patients (including lung, melanoma, breast, renal, and colorectal cancer patients) develop brain metastases. Current systemic therapies often show limited efficacy for brain metastasis treatment and support palliative care only. In this preclinical study, short-term cultured primary cancer cells from brain metastasis patients were molecularly characterized using next-generation sequencing and functionally evaluated using high-throughput in vitro drug screening to identify pharmacological treatment sensitivities. Next-generation sequencing identified matched genetic alterations between brain metastasis tissues and corresponding patient-derived cultured cancer cells. Mutant allele frequencies indicated that short-term cultures of cancer cells reflect the existence of distinct subpopulations of cancer cells within brain metastases. This indicates that our short-term cultured primary brain metastasis cells are suitable models for recapitulating heterogeneity in the sensitivity of tumor cells to anti-cancer drugs. Employing our in-house high-throughput in vitro drug screening platform, we screened the primary cultures from the brain metastasis patients for response to 267 anticancer compounds, and integrated the molecular and high-throughput drug screening data sets. Our preclinical study provides proof-of-concept for combining molecular profiling with in vitro drug screening for improved prediction of therapeutic vulnerabilities in brain metastasis patients. This approach could advance the use of patient-derived cancer cells in clinical practice and might eventually facilitate decision-making for personalized drug treatment.

Machine learning and drug discovery for neglected tropical diseases

Neglected tropical diseases affect millions of individuals and cause loss of productivity worldwide. They are common in developing countries without the financial resources for research and drug development. With increased availability of data from high throughput screening, machine learning has been introduced into the drug discovery process. Models can be trained to predict biological activities of compounds before working in the lab. In this study, we use three publicly available, high-throughput screening datasets to train machine learning models to predict biological activities related to inhibition of species that cause leishmaniasis, American trypanosomiasis (Chagas disease), and African trypanosomiasis (sleeping sickness). We compare machine learning models (tree based models, naive Bayes classifiers, and neural networks), featurizing methods (circular fingerprints, MACCS fingerprints, and RDKit descriptors), and techniques to deal with the imbalanced data (oversampling, undersampling, class weight/sample weight).

Predicting activatory and inhibitory drug-target interactions based on structural compound representations and genetically perturbed transcriptomes.

A computational approach to identifying drug-target interactions (DTIs) is a credible strategy for accelerating drug development and understanding the mechanisms of action of small molecules. However, current methods to predict DTIs have mainly focused on identifying simple interactions, requiring further experiments to understand mechanism of drug. Here, we propose AI-DTI, a novel method that predicts activatory and inhibitory DTIs by combining the mol2vec and genetically perturbed transcriptomes. We trained the model on large-scale DTIs with MoA and found that our model outperformed a previous model that predicted activatory and inhibitory DTIs. Data augmentation of target feature vectors enabled the model to predict DTIs for a wide druggable targets. Our method achieved substantial performance in an independent dataset where the target was unseen in the training set and a high-throughput screening dataset where positive and negative samples were explicitly defined. Also, our method successfully rediscovered approximately half of the DTIs for drugs used in the treatment of COVID-19. These results indicate that AI-DTI is a practically useful tool for guiding drug discovery processes and generating plausible hypotheses that can reveal unknown mechanisms of drug action.

A Multi-Objective Active Learning Platform and Web App for Reaction Optimization.

We report the development of an open-source experimental design via Bayesian optimization platform for multi-objective reaction optimization. Using high-throughput experimentation (HTE) and virtual screening data sets containing high-dimensional continuous and discrete variables, we optimized the performance of the platform by fine-tuning the algorithm components such as reaction encodings, surrogate model parameters, and initialization techniques. Having established the framework, we applied the optimizer to real-world test scenarios for the simultaneous optimization of the reaction yield and enantioselectivity in a Ni/photoredox-catalyzed enantioselective cross-electrophile coupling of styrene oxide with two different aryl iodide substrates. Starting with no previous experimental data, the Bayesian optimizer identified reaction conditions that surpassed the previously human-driven optimization campaigns within 15 and 24 experiments, for each substrate, among 1728 possible configurations available in each optimization. To make the platform more accessible to nonexperts, we developed a graphical user interface (GUI) that can be accessed online through a web-based application and incorporated features such as condition modification on the fly and data visualization. This web application does not require software installation, removing any programming barrier to use the platform, which enables chemists to integrate Bayesian optimization routines into their everyday laboratory practices.

Gene expression based inference of cancer drug sensitivity

Inter and intra-tumoral heterogeneity are major stumbling blocks in the treatment of cancer and are responsible for imparting differential drug responses in cancer patients. Recently, the availability of high-throughput screening datasets has paved the way for machine learning based personalized therapy recommendations using the molecular profiles of cancer specimens. In this study, we introduce Precily, a predictive modeling approach to infer treatment response in cancers using gene expression data. In this context, we demonstrate the benefits of considering pathway activity estimates in tandem with drug descriptors as features. We apply Precily on single-cell and bulk RNA sequencing data associated with hundreds of cancer cell lines. We then assess the predictability of treatment outcomes using our in-house prostate cancer cell line and xenografts datasets exposed to differential treatment conditions. Further, we demonstrate the applicability of our approach on patient drug response data from The Cancer Genome Atlas and an independent clinical study describing the treatment journey of three melanoma patients. Our findings highlight the importance of chemo-transcriptomics approaches in cancer treatment selection.

High-throughput screening and literature data-driven machine learning-assisted investigation of multi-component La2O3-based catalysts for the oxidative coupling of methane

Unique 30 types of multi-component La2O3-based catalysts for oxidative coupling of methane were discovered in 75 types of selected catalysts based on high-throughput screening and literature datasets with multi-output machine learning approaches.

Literature-based translation from synthetic lethality screening into therapeutics targets: CD82 is a novel target for KRAS mutation in colon cancer

Synthetic lethality (SL) is an emerging therapeutic paradigm in cancer. We introduced a different approach to prioritize SL gene pairs through literature mining and RAS-mutant high-throughput screening (HTS) data. We matched essential genes from text-mining and mutant genes from the COSMIC and CCLE HTS datasets to build a prediction model of SL gene pairs. CCLE gene expression data were used to enrich the essential-mutant SL gene pairs using Spearman’s correlation coefficient and literature mining. In total, 223 essential trigger terms were extracted and ranked. The threshold of the essential gene score (Sg) was set to 10. We identified 586 genes essential for the SL prediction model of colon cancer. Seven essential RAS-mutant SL gene pairs were identified in our model, including CD82-KRAS/NRAS, PEBP1-NRAS, MT-CO2-HRAS, IFI27-NRAS/KRAS, and SUMO1-HRAS gene pairs. Using RAS-mutant HTS data validation, we identified two potential SL gene pairs, including the CD82 (essential gene)–KRAS (mutant gene) pair and CD82–NRAS pair in the DLD-1 colon cancer cell line (Spearman’s correlation p-values = 0.004786 and 0.00249, respectively). Based on further annotations by PubChem, we observed that digitonin targeted the complex comprising CD82, especially in KRAS-mutated HCT116 cancer cells. Moreover, we experimentally demonstrated that CD82 exhibited selective vulnerability in KRAS-mutant colorectal cancer. We used literature mining and HTS data to identify candidates for SL targets for RAS-mutant colon cancer.

Identification and characterization of a fungal-selective glutaminyl tRNA synthetase inhibitor with potent activity against Candida albicans

Candida albicans is the leading cause of systemic candidiasis. Effective treatment is threatened by a dearth of antifungal options and the emergence of resistance. Thus, there is an urgent need to identify novel therapeutic targets to expand our antifungal armamentarium. A promising approach is the discovery of essential genes, as most antimicrobials target essential bioprocesses. Despite detailed characterization of gene essentiality in Saccharomyces cerevisiae,defining essential targets in the pathogen of interest is necessary due to the high level of divergence between these organisms. Thus, using a machine learning algorithm we generated a comprehensive prediction of all genes essential in C. albicans. We leveraged our essentiality predictions with high-throughput screening and chemogenomic datasets to assign the mechanism of action of a previously uncharacterized compound. We identified T-035897 as a molecule with potent bioactivity against C. albicans. Prior chemogenomic profiling in S. cerevisiae suggested that T-035897 targets the glutaminyl tRNA synthetase Gln4, whose homolog in C. albicans was predicted and verified to be required for viability. To confirm the mechanism of T-035897 in C. albicans, we performed haploinsufficiency profiling,which supported Gln4as the target. In parallel, selection of resistant mutants and targeted sequencing uncovered substitutions in the Gln4 catalytic domain. Moreover, T-035897 inhibited translation in afluorescence-based reporter assay. Finally, T-035897 selectively abrogated fungal cell growth in a co-culture model with mammalian cells. Thus, we highlight the power of leveraging essentiality datasets in order to characterize compounds with potent antifungal activity in an effort to unveil novel therapeutic strategies.

Predictive Models to Identify Small Molecule Activators and Inhibitors of Opioid Receptors.

Opioid receptors (OPRs) are the main targets for the treatment of pain and related disorders. The opiate compounds that activate these receptors are effective analgesics but their use leads to adverse effects, and they often are highly addictive drugs of abuse. There is an urgent need for alternative chemicals that are analgesics and to reduce/avoid the unwanted effects in order to relieve the public health crisis of opioid addiction. Here, we aim to develop computational models to predict the OPR activity of small molecule compounds based on chemical structures and apply these models to identify novel OPR active compounds. We used four different machine learning algorithms to build models based on quantitative high throughput screening (qHTS) data sets of three OPRs in both agonist and antagonist modes. The best performing models were applied to virtually screen a large collection of compounds. The model predicted active compounds were experimentally validated using the same qHTS assays that generated the training data. Random forest was the best classifier with the highest performance metrics, and the mu OPR (OPRM)-agonist model achieved the best performance measured by AUC-ROC (0.88) and MCC (0.7) values. The model predicted actives resulted in hit rates ranging from 2.3% (delta OPR-agonist) to 15.8% (OPRM-agonist) after experimental confirmation. Compared to the original assay hit rate, all models enriched the hit rate by ≥2-fold. Our approach produced robust OPR prediction models that can be applied to prioritize compounds from large libraries for further experimental validation. The models identified several novel potent compounds as activators/inhibitors of OPRs that were confirmed experimentally. The potent hits were further investigated using molecular docking to find the interactions of the novel ligands in the active site of the corresponding OPR.

FutureTox IV Workshop Summary: Predictive Toxicology for Healthy Children.

FutureTox IV, a Society of Toxicology Contemporary Concepts in Toxicology workshop, was held in November 2018. Building upon FutureTox I, II, and III, this conference focused on the latest science and technology for in vitro profiling and in silico modeling as it relates to predictive developmental and reproductive toxicity (DART). Publicly available high-throughput screening data sets are now available for broad in vitro profiling of bioactivities across large inventories of chemicals. Coupling this vast amount of mechanistic data with a deeper understanding of molecular embryology and post-natal development lays the groundwork for using new approach methodologies (NAMs) to evaluate chemical toxicity, drug efficacy, and safety assessment for embryo-fetal development. NAM is a term recently adopted in reference to any technology, methodology, approach, or combination thereof that can be used to provide information on chemical hazard and risk assessment to avoid the use of intact animals (U.S. Environmental Protection Agency [EPA], Strategic plan to promote the development and implementation of alternative test methods within the tsca program, 2018, https://www.epa.gov/sites/production/files/2018-06/documents/epa_alt_strat_plan_6-20-18_clean_final.pdf). There are challenges to implementing NAMs to evaluate chemicals for developmental toxicity compared with adult toxicity. This forum article reviews the 2018 workshop activities, highlighting challenges and opportunities for applying NAMs for adverse pregnancy outcomes (eg, preterm labor, malformations, low birth weight) as well as disorders manifesting postnatally (eg, neurodevelopmental impairment, breast cancer, cardiovascular disease, fertility). DART is an important concern for different regulatory statutes and test guidelines. Leveraging advancements in such approaches and the accompanying efficiencies to detecting potential hazards to human development are the unifying concepts toward implementing NAMs in DART testing. Although use of NAMs for higher level regulatory decision making is still on the horizon, the conference highlighted novel testing platforms and computational models that cover multiple levels of biological organization, with the unique temporal dynamics of embryonic development, and novel approaches for estimating toxicokinetic parameters essential in supporting in vitro to in vivo extrapolation.

Statistical models for identifying frequent hitters in high throughput screening

High throughput screening (HTS) interrogates compound libraries to find those that are “active” in an assay. To better understand compound behavior in HTS, we assessed an existing binomial survivor function (BSF) model of “frequent hitters” using 872 publicly available HTS data sets. We found large numbers of “infrequent hitters” using this model leading us to reject the BSF for identifying “frequent hitters.” As alternatives, we investigated generalized logistic, gamma, and negative binomial distributions as models for compound behavior. The gamma model reduced the proportion of both frequent and infrequent hitters relative to the BSF. Within this data set, conclusions about individual compound behavior were limited by the number of times individual compounds were tested (1–1613 times) and disproportionate testing of some compounds. Specifically, most tests (78%) were on a 309,847-compound subset (17.6% of compounds) each tested ≥ 300 times. We concluded that the disproportionate retesting of some compounds represents compound repurposing at scale rather than drug discovery. The approach to drug discovery represented by these 872 data sets characterizes the assays well by challenging them with many compounds while each compound is characterized poorly with a single assay. Aggregating the testing information from each compound across the multiple screens yielded a continuum with no clear boundary between normal and frequent hitting compounds.

Inductive transfer learning for molecular activity prediction: Next-Gen QSAR Models with MolPMoFiT

Deep neural networks can directly learn from chemical structures without extensive, user-driven selection of descriptors in order to predict molecular properties/activities with high reliability. But these approaches typically require large training sets to learn the endpoint-specific structural features and ensure reasonable prediction accuracy. Even though large datasets are becoming the new normal in drug discovery, especially when it comes to high-throughput screening or metabolomics datasets, one should also consider smaller datasets with challenging endpoints to model and forecast. Thus, it would be highly relevant to better utilize the tremendous compendium of unlabeled compounds from publicly-available datasets for improving the model performances for the user’s particular series of compounds. In this study, we propose the Molecular Prediction Model Fine-Tuning (MolPMoFiT) approach, an effective transfer learning method based on self-supervised pre-training + task-specific fine-tuning for QSPR/QSAR modeling. A large-scale molecular structure prediction model is pre-trained using one million unlabeled molecules from ChEMBL in a self-supervised learning manner, and can then be fine-tuned on various QSPR/QSAR tasks for smaller chemical datasets with specific endpoints. Herein, the method is evaluated on four benchmark datasets (lipophilicity, FreeSolv, HIV, and blood–brain barrier penetration). The results showed the method can achieve strong performances for all four datasets compared to other state-of-the-art machine learning modeling techniques reported in the literature so far.

Ranking Molecules with Vanishing Kernels and a Single Parameter: Active Applicability Domain Included.

In ligand-based virtual screening, high-throughput screening (HTS) data sets can be exploited to train classification models. Such models can be used to prioritize yet untested molecules, from the most likely active (against a protein target of interest) to the least likely active. In this study, a single-parameter ranking method with an Applicability Domain (AD) is proposed. In effect, Kernel Density Estimates (KDE) are revisited to improve their computational efficiency and incorporate an AD. Two modifications are proposed: (i) using vanishing kernels (i.e., kernel functions with a finite support) and (ii) using the Tanimoto distance between molecular fingerprints as a radial basis function. This construction is termed "Vanishing Ranking Kernels" (VRK). Using VRK on 21 HTS assays, it is shown that VRK can compete in performance with a graph convolutional deep neural network. VRK are conceptually simple and fast to train. During training, they require optimizing a single parameter. A trained VRK model usually defines an active AD. Exploiting this AD can significantly increase the screening frequency of a VRK model. Software: https://github.com/UnixJunkie/rankers. Data sets: https://zenodo.org/record/1320776 and https://zenodo.org/record/3540423.

The Distribution of Standard Deviations Applied to High Throughput Screening

High throughput screening (HTS) assesses compound libraries for “activity” using target assays. A subset of HTS data contains a large number of sample measurements replicated a small number of times providing an opportunity to introduce the distribution of standard deviations (DSD). Applying the DSD to some HTS data sets revealed signs of bias in some of the data and discovered a sub-population of compounds exhibiting high variability which may be difficult to screen. In the data examined, 21% of 1189 such compounds were pan-assay interference compounds. This proportion reached 57% for the most closely related compounds within the sub-population. Using the DSD, large HTS data sets can be modelled in many cases as two distributions: a large group of nearly normally distributed “inactive” compounds and a residual distribution of “active” compounds. The latter were not normally distributed, overlapped inactive distributions – on both sides –, and were larger than typically assumed. As such, a large number of compounds are being misclassified as “inactive” or are invisible to current methods which could become the next generation of drugs. Although applied here to HTS, it is applicable to data sets with a large number of samples measured a small number of times.

A Distance-Based Boolean Applicability Domain for Classification of High Throughput Screening Data.

In Quantitative Structure-Activity Relationship (QSAR) modeling, one must come up with an activity model but also with an applicability domain for that model. Some existing methods to create an applicability domain are complex, hard to implement, and/or difficult to interpret. Also, they often require the user to select a threshold value, or they embed an empirical constant. In this work, we propose a trivial to interpret and fully automatic Distance-Based Boolean Applicability Domain (DBBAD) algorithm for category QSAR. In retrospective experiments on High Throughput Screening data sets, this applicability domain improves the classification performance and early retrieval of support vector machine and random forest based classifiers, while improving the scaffold diversity among top-ranked active molecules.

Uncertainty quantification in ToxCast high throughput screening.

High throughput screening (HTS) projects like the U.S. Environmental Protection Agency's ToxCast program are required to address the large and rapidly increasing number of chemicals for which we have little to no toxicity measurements. Concentration-response parameters such as potency and efficacy are extracted from HTS data using nonlinear regression, and models and analyses built from these parameters are used to predict in vivo and in vitro toxicity of thousands of chemicals. How these predictions are impacted by uncertainties that stem from parameter estimation and propagated through the models and analyses has not been well explored. While data size and complexity makes uncertainty quantification computationally expensive for HTS datasets, continued advancements in computational resources have allowed these computational challenges to be met. This study uses nonparametric bootstrap resampling to calculate uncertainties in concentration-response parameters from a variety of HTS assays. Using the ToxCast estrogen receptor model for bioactivity as a case study, we highlight how these uncertainties can be propagated through models to quantify the uncertainty in model outputs. Uncertainty quantification in model outputs is used to identify potential false positives and false negatives and to determine the distribution of model values around semi-arbitrary activity cutoffs, increasing confidence in model predictions. At the individual chemical-assay level, curves with high variability are flagged for manual inspection or retesting, focusing subject-matter-expert time on results that need further input. This work improves the confidence of predictions made using HTS data, increasing the ability to use this data in risk assessment.

Data Mining and Computational Modeling of High-Throughput Screening Datasets.

We are now seeing the benefit of investments made over the last decade in high-throughput screening (HTS) that is resulting in large structure activity datasets entering public and open databases such as ChEMBL and PubChem. The growth of academic HTS screening centers and the increasing move to academia for early stage drug discovery suggests a great need for the informatics tools and methods to mine such data and learn from it. Collaborative Drug Discovery, Inc. (CDD) has developed a number of tools for storing, mining, securely and selectively sharing, as well as learning from such HTS data. We present a new web based data mining and visualization module directly within the CDD Vault platform for high-throughput drug discovery data that makes use of a novel technology stack following modern reactive design principles. We also describe CDD Models within the CDD Vault platform that enables researchers to share models, share predictions from models, and create models from distributed, heterogeneous data. Our system is built on top of the Collaborative Drug Discovery Vault Activity and Registration data repository ecosystem which allows users to manipulate and visualize thousands of molecules in real time. This can be performed in any browser on any platform. In this chapter we present examples of its use with public datasets in CDD Vault. Such approaches can complement other cheminformatics tools, whether open source or commercial, in providing approaches for data mining and modeling of HTS data.