True Binders Research Articles

TPepRet: a deep learning model for characterizing T cell receptors-antigen binding patterns.

T-cell receptors (TCRs) elicit and mediate the adaptive immune response by recognizing antigenic peptides, a process pivotal for cancer immunotherapy, vaccine design, and autoimmune disease management. Understanding the intricate binding patterns between TCRs and peptides is critical for advancing these clinical applications. While several computational tools have been developed, they neglect the directional semantics inherent in sequence data, which are essential for accurately characterizing TCR-peptide interactions. To address this gap, we develop TPepRet, an innovative model that integrates subsequence mining with semantic integration capabilities. TPepRet combines the strengths of the Bidirectional Gated Recurrent Unit (BiGRU) network for capturing bidirectional sequence dependencies with the Large Language Model framework to analyze subsequences and global sequences comprehensively, which enables TPepRet to accurately decipher the semantic binding relationship between TCRs and peptides. We have evaluated TPepRet to a range of challenging scenarios, including performance benchmarking against other tools using diverse datasets, analysis of peptide binding preferences, characterization of T cells clonal expansion, identification of true binder in complex environments, assessment of key binding sites through alanine scanning, validation against expression rates from large-scale datasets, and ability to screen SARS-CoV-2 TCRs. The comprehensive results suggest that TPepRet outperforms existing tools. We believe TPepRet will become an effective tool for understanding TCR-peptide binding in clinical treatment. The source code can be obtained from https://github.com/CSUBioGroup/TPepRet.git. Supplementary data are available at Bioinformatics online.

Jointly Embedding Protein Structures and Sequences through Residue Level Alignment

The relationships between protein sequences, structures, and functions are determined by complex codes that scientists aim to decipher. While structures contain key information about proteins' biochemical functions, they are often experimentally difficult to obtain. In contrast, protein sequences are abundant but are a step removed from function. In this paper, we propose residue level alignment (RLA)—a self-supervised objective for aligning sequence and structure embedding spaces. By situating sequence and structure encoders within the same latent space, RLA enriches the sequence encoder with spatial information. Moreover, our framework enables us to measure the similarity between a sequence and structure by comparing their RLA embeddings. We show how RLA similarity scores can be used for binder design by selecting true binders from sets of designed binders. RLA scores are informative even when they are calculated given only the backbone structure of the binder and no binder sequence information, which simulates the information available in many early-stage binder design libraries. RLA performs similarly to benchmark methods and is orders of magnitude faster, making it a valuable new screening tool for binder design pipelines. Published by the American Physical Society 2024

Evaluating NetMHCpan performance on non-European HLA alleles not present in training data.

Bias in neural network model training datasets has been observed to decrease prediction accuracy for groups underrepresented in training data. Thus, investigating the composition of training datasets used in machine learning models with healthcare applications is vital to ensure equity. Two such machine learning models are NetMHCpan-4.1 and NetMHCIIpan-4.0, used to predict antigen binding scores to major histocompatibility complex class I and II molecules, respectively. As antigen presentation is a critical step in mounting the adaptive immune response, previous work has used these or similar predictions models in a broad array of applications, from explaining asymptomatic viral infection to cancer neoantigen prediction. However, these models have also been shown to be biased toward hydrophobic peptides, suggesting the network could also contain other sources of bias. Here, we report the composition of the networks' training datasets are heavily biased toward European Caucasian individuals and against Asian and Pacific Islander individuals. We test the ability of NetMHCpan-4.1 and NetMHCpan-4.0 to distinguish true binders from randomly generated peptides on alleles not included in the training datasets. Unexpectedly, we fail to find evidence that the disparities in training data lead to a meaningful difference in prediction quality for alleles not present in the training data. We attempt to explain this result by mapping the HLA sequence space to determine the sequence diversity of the training dataset. Furthermore, we link the residues which have the greatest impact on NetMHCpan predictions to structural features for three alleles (HLA-A*34:01, HLA-C*04:03, HLA-DRB1*12:02).

Machine-Learning-Based Data Analysis Method for Cell-Based Selection of DNA-Encoded Libraries.

DNA-encoded library (DEL) is a powerful ligand discovery technology that has been widely adopted in the pharmaceutical industry. DEL selections are typically performed with a purified protein target immobilized on a matrix or in solution phase. Recently, DELs have also been used to interrogate the targets in the complex biological environment, such as membrane proteins on live cells. However, due to the complex landscape of the cell surface, the selection inevitably involves significant nonspecific interactions, and the selection data are much noisier than the ones with purified proteins, making reliable hit identification highly challenging. Researchers have developed several approaches to denoise DEL datasets, but it remains unclear whether they are suitable for cell-based DEL selections. Here, we report the proof-of-principle of a new machine-learning (ML)-based approach to process cell-based DEL selection datasets by using a Maximum A Posteriori (MAP) estimation loss function, a probabilistic framework that can account for and quantify uncertainties of noisy data. We applied the approach to a DEL selection dataset, where a library of 7,721,415 compounds was selected against a purified carbonic anhydrase 2 (CA-2) and a cell line expressing the membrane protein carbonic anhydrase 12 (CA-12). The extended-connectivity fingerprint (ECFP)-based regression model using the MAP loss function was able to identify true binders and also reliable structure-activity relationship (SAR) from the noisy cell-based selection datasets. In addition, the regularized enrichment metric (known as MAP enrichment) could also be calculated directly without involving the specific machine-learning model, effectively suppressing low-confidence outliers and enhancing the signal-to-noise ratio. Future applications of this method will focus on de novo ligand discovery from cell-based DEL selections.

DyScore: A Boosting Scoring Method with Dynamic Properties for Identifying True Binders and Nonbinders in Structure-Based Drug Discovery.

The accurate prediction of protein-ligand binding affinity is critical for the success of computer-aided drug discovery. However, the accuracy of current scoring functions is usually unsatisfactory due to their rough approximation or sometimes even omittance of many factors involved in protein-ligand binding. For instance, the intrinsic dynamics of the protein-ligand binding state is usually disregarded in scoring function because these rapid binding affinity prediction approaches are only based on a representative complex structure of the protein and ligand in the binding state. That is, the dynamic protein-ligand binding complex ensembles are simplified as a static snapshot in calculation. In this study, two novel features were proposed for characterizing the dynamic properties of protein-ligand binding based on the static structure of the complex, which is expected to be a valuable complement to the current scoring functions. The two features demonstrate the geometry-shape matching between a protein and a ligand as well as the dynamic stability of protein-ligand binding. We further combined these two novel features with several classical scoring functions to develop a binary classification model called DyScore that uses the Extreme Gradient Boosting algorithm to classify compound poses as binders or non-binders. We have found that DyScore achieves state-of-the-art performance in distinguishing active and decoy ligands on both enhanced DUD data set and external test sets with both proposed novel features showing significant contributions to the improved performance. Especially, DyScore exhibits superior performance on early recognition, a crucial requirement for success in virtual screening and de novo drug design. The standalone version of DyScore and Dyscore-MF are freely available to all at: https://github.com/YanjunLi-CS/dyscore.

A cell-free nanobody engineering platform rapidly generates SARS-CoV-2 neutralizing nanobodies

Antibody engineering technologies face increasing demands for speed, reliability and scale. We develop CeVICA, a cell-free nanobody engineering platform that uses ribosome display for in vitro selection of nanobodies from a library of 1011 randomized sequences. We apply CeVICA to engineer nanobodies against the Receptor Binding Domain (RBD) of SARS-CoV-2 spike protein and identify >800 binder families using a computational pipeline based on CDR-directed clustering. Among 38 experimentally-tested families, 30 are true RBD binders and 11 inhibit SARS-CoV-2 pseudotyped virus infection. Affinity maturation and multivalency engineering increase nanobody binding affinity and yield a virus neutralizer with picomolar IC50. Furthermore, the capability of CeVICA for comprehensive binder prediction allows us to validate the fitness of our nanobody library. CeVICA offers an integrated solution for rapid generation of divergent synthetic nanobodies with tunable affinities in vitro and may serve as the basis for automated and highly parallel nanobody engineering.

True Accuracy of Fast Scoring Functions to Predict High-Throughput Screening Data from Docking Poses: The Simpler the Better.

Hundreds of fast scoring functions have been developed over the last 20 years to predict binding free energies from three-dimensional structures of protein-ligand complexes. Despite numerous statistical promises, we believe that none of them has been properly validated for daily prospective high-throughput virtual screening studies, mostly because in silico screening challenges usually employ artificially built and biased datasets. We here carry out a fully unbiased evaluation of four scoring functions (Pafnucy, ΔvinaRF20, IFP, and GRIM) on an in-house developed data collection of experimental high-confidence screening data (LIT-PCBA) covering about 3 million data points on 15 diverse pharmaceutical targets. All four scoring functions were applied to rescore the docking poses of LIT-PCBA compounds in conditions mimicking exactly standard drug discovery scenarios and were compared in terms of propensity to enrich true binders in the top 1%-ranked hit lists. Interestingly, rescoring based on simple interaction fingerprints or interaction graphs outperforms state-of-the-art machine learning and deep learning scoring functions in most of the cases. The current study notably highlights the strong tendency of deep learning methods to predict affinity values within a very narrow range centered on the mean value of samples used for training. Moreover, it suggests that knowledge of pre-existing binding modes is the key to detecting the most potent binders.

Toward the assembly and characterization of an encoded library hit confirmation platform: Bead-Assisted Ligand Isolation Mass Spectrometry (BALI-MS)

The ability to predict chemical structure from DNA sequence has to date been a necessary cornerstone of DNA-encoded library technology. DNA-encoded libraries (DELs) are typically screened by immobilized affinity selection and enriched library members are identified by counting the number of times an individual compound’s sequence is observed in the resultant dataset. Those with high signal reads (DEL hits) are subsequently followed up through off-DNA synthesis of the predicted small molecule structures. However, hits followed-up in this manner often fail to translate to confirmed ligands. To address this low conversion rate of DEL hits to off-DNA ligands, we have developed an approach that eliminates the reliance on chemical structure prediction from DNA sequence. Here we describe our method of combining non-combinatorial resynthesis on-DNA following library procedures as a rapid means to assess the probable molecules attached to the DNA barcode. Furthermore, we apply our Bead-Assisted Ligand Isolation Mass Spectrometry (BALI-MS) technique to identify the true binders found within the mixtures of on-DNA synthesis products. Finally, we describe a Normalized Enrichment (NE) metric that allows for the quantitative assessment of affinity selection in these studies. We exemplify how this combined approach enables the identification of putative hit matter against a clinically relevant therapeutic target bisphosphoglycerate mutase, BPGM.

Molecular modeling study of the testosterone metabolizing enzyme UDP-glucuronosyltransferase 2B17

The dominant sex hormone testosterone is mainly metabolized by liver enzymes belonging to the uridine-diphospho (UDP) glucuronosyltransferase (UGT) family. These enzymes are the main phase II enzymes, and they have an important role in the detoxification of endogenous and exogenous compounds in humans. The aim of the present study was to improve the understanding of the binding properties of UGT2B17. A homology modelling procedure was used to generate models of the UGT2B17 enzyme based on templates with known crystal structures. Molecular docking of inhibitors was performed to gain further insights in the interactions between ligand and binding site, and to determine which of the models had the best accuracy. ROC curves were made to evaluate the ability of the models to differentiate between binders (inhibitors) and non-binders (decoys). When comparing the four models, which were based on four different crystal structures, the model based on the 4AMG crystal structure was the most accurate in distinguishing between true binders and non-binders. Investigating pharmacological UGT2B17 inhibition may provide novel treatment for patients with low testosterone levels. Such treatment may elevate endogenous testosterone levels and provide a more predictable increase in serum concentrations rather than un-physiological elevation of serum levels through direct treatment with testosterone, and this could be favorable both for giving a predictable treatment regime with reduced chances of serious adverse effects. The present study may serve as a tool in the search for novel drugs aiming for increasing testosterone levels.

LigGrep: a tool for filtering docked poses to improve virtual-screening hit rates

Structure-based virtual screening (VS) uses computer docking to prioritize candidate small-molecule ligands for subsequent experimental testing. Docking programs evaluate molecular binding in part by predicting the geometry with which a given compound might bind a target receptor (e.g., the docked “pose” relative to a protein target). Candidate ligands predicted to participate in the same intermolecular interactions typical of known ligands (or ligands that bind related proteins) are arguably more likely to be true binders. Some docking programs allow users to apply constraints during the docking process with the goal of prioritizing these critical interactions. But these programs often have restrictive and/or expensive licenses, and many popular open-source docking programs (e.g., AutoDock Vina) lack this important functionality. We present LigGrep, a free, open-source program that addresses this limitation. As input, LigGrep accepts a protein receptor file, a directory containing many docked-compound files, and a list of user-specified filters describing critical receptor/ligand interactions. LigGrep evaluates each docked pose and outputs the names of the compounds with poses that pass all filters. To demonstrate utility, we show that LigGrep can improve the hit rates of test VS targeting H. sapiens poly(ADPribose) polymerase 1 (HsPARP1), H. sapiens peptidyl-prolyl cis-trans isomerase NIMA-interacting 1 (HsPin1p), and S. cerevisiae hexokinase-2 (ScHxk2p). We hope that LigGrep will be a useful tool for the computational biology community. A copy is available free of charge at http://durrantlab.com/liggrep/.

An ABSINTH-Based Protocol for Predicting Binding Affinities between Proteins and Small Molecules.

The core task in computational drug discovery is to accurately predict binding free energies in receptor-ligand systems for large libraries of putative binders. Here, the ABSINTH implicit solvent model and force field are extended to describe small, organic molecules and their interactions with proteins. We show that an automatic pipeline based on partitioning arbitrary molecules into substructures corresponding to model compounds with known free energies of solvation can be combined with the CHARMM general force field into a method that is successful at the two important challenges a scoring function faces in virtual screening work flows: it ranks known binders with correlation values rivaling that of comparable state-of-the-art methods and it enriches true binders in a set of decoys. Our protocol introduces innovative modifications to common virtual screening workflows, notably the use of explicit ions as competitors and the integration over multiple protein and ligand species differing in their protonation states. We demonstrate the value of modifications to both the protocol and ABSINTH itself. We conclude by discussing the limitations of high-throughput implicit methods such as the one proposed here.

Denoising DNA Encoded Library Screens with Sparse Learning.

DNA-encoded libraries (DELs) are large, pooled collections of compounds in which every library member is attached to a stretch of DNA encoding its complete synthetic history. DEL-based hit discovery involves affinity selection of the library against a protein of interest, whereby compounds retained by the target are subsequently identified by next-generation sequencing of the corresponding DNA tags. When analyzing the resulting data, one typically assumes that sequencing output (i.e., read counts) is proportional to the binding affinity of a given compound, thus enabling hit prioritization and elucidation of any underlying structure-activity relationships (SAR). This assumption, though, tends to be severely confounded by a number of factors, including variable reaction yields, presence of incomplete products masquerading as their intended counterparts, and sequencing noise. In practice, these confounders are often ignored, potentially contributing to low hit validation rates, and universally leading to loss of valuable information. To address this issue, we have developed a method for comprehensively denoising DEL selection outputs. Our method, dubbed "deldenoiser", is based on sparse learning and leverages inputs that are commonly available within a DEL generation and screening workflow. Using simulated and publicly available DEL affinity selection data, we show that "deldenoiser" is not only able to recover and rank true binders much more robustly than read count-based approaches but also that it yields scores, which accurately capture the underlying SAR. The proposed method can, thus, be of significant utility in hit prioritization following DEL screens.

Isolation of Monoclonal Antibodies from Zika Virus-Infected Patient Samples.

The combination of sorting antigen-specific memory B cells with determining immunoglobulin (Ig) genes at the single-cell level enables the isolation of monoclonal antibodies (mAbs) in individuals. This method requires a small amount of blood (usually 10 mL) and is rapid (less than 2 weeks to isolate antigen-specific mAbs). Due to the application of antigens as the bait to capture the specific memory B cells, the majority of isolated mAbs are true binders to the antigen, which increases the isolation efficiency. Here, applying this approach, we describe the characterization of mAbs against Zika virus from a convalescent patient sample. From 10 mL whole blood, we sorted 33 Zika envelope (E) protein-interacting single memory B cells. The Ig genes from 15 cells were determined, and 13 mAbs were found that bind to Zika E protein with varied binding affinities.

Molecular modeling and docking studies of the oxytocin receptor.

Low oxytocin (OT) level is involved in a number of psychiatric diseases, indicating that OT could be used to aid treating these disorders. OT itself is unable to cross the blood-brain barrier, and development of new small nonpeptide drugs targeting the OT receptor (OXTR) may be beneficial for treating mental disorders. Results & methodology: Three OXTR models were constructed based on crystallized homologous proteins (Protein Data Bank [PDB]: 2Y00, PDB: 4BVN and PDB: 4LDE). The abilities of the models to discriminate between true binders and decoys were analyzed using receiver operating characteristics curves, and the 4LDE-based model gave the best result. The present study demonstrates that the 4LDE-based model may be suitable as a tool for the development of novel drugs targeting OXTR.

LigQ: A Webserver to Select and Prepare Ligands for Virtual Screening.

Virtual screening is a powerful methodology to search for new small molecule inhibitors against a desired molecular target. Usually, it involves evaluating thousands of compounds (derived from large databases) in order to select a set of potential binders that will be tested in the wet-lab. The number of tested compounds is directly proportional to the cost, and thus, the best possible set of ligands is the one with the highest number of true binders, for the smallest possible compound set size. Therefore, methods that are able to trim down large universal data sets enriching them in potential binders are highly appreciated. Here we present LigQ, a free webserver that is able to (i) determine best structure and ligand binding pocket for a desired protein, (ii) find known binders, as well as potential ligands known to bind to similar protein domains, (iii) most importantly, select a small set of commercial compounds enriched in potential binders, and (iv) prepare them for virtual screening. LigQ was tested with several proteins, showing an impressive capacity to retrieve true ligands from large data sets, achieving enrichment factors of over 10%. LigQ is available at http://ligq.qb.fcen.uba.ar/ .

Phage display biopanning and isolation of target-unrelated peptides: in search of nonspecific binders hidden in a combinatorial library.

Phage display is known as a powerful methodology for the identification of targeting ligands that specifically bind to a variety of targets. The high-throughput screening of phage display combinatorial peptide libraries is performed through the affinity selection method of biopanning. Although phage display selection has proven very successful in the discovery of numerous high-affinity target-binding peptides with potential application in drug discovery and delivery, the enrichment of false-positive target-unrelated peptides (TUPs) without any actual affinity towards the target remains a major problem of library screening. Selection-related TUPs may emerge because of binding to the components of the screening system rather than the target. Propagation-related TUPs may arise as a result of faster growth rate of some phage clones enabling them to outcompete slow-propagating clones. Amplification of the library between rounds of biopanning makes a significant contribution to the selection of phage clones with propagation advantage. Distinguishing nonspecific TUPs from true target binders is of particular importance for the translation of biopanning findings from basic research to clinical applications. Different experimental and in silico approaches are applied to assess the specificity of phage display-derived peptides towards the target. Bioinformatic tools are playing a rapidly growing role in the analysis of biopanning data and identification of target-irrelevant TUPs. Recent progress in the introduction of efficient strategies for TUP detection holds enormous promise for the discovery of clinically relevant cell- and tissue-homing peptides and paves the way for the development of novel targeted diagnostic and therapeutic platforms in pharmaceutical areas.

A polystyrene binding target-unrelated peptide isolated in the screening of phage display library

Phage display is a powerful methodology for the identification of peptide ligands binding to any desired target. However, the selection of target-unrelated peptides (TUPs) appears as a huge problem in the screening of phage display libraries through biopanning. The phage-displayed peptide TLHPAAD has been isolated both in our laboratory and by another reserach group on completely different screening targets prompting us to hypothesize that it may be a potential TUP. In the current study, we analyzed the binding characteristics and propagation rate of phage clone displaying TLHPAAD peptide (SW-TUP clone). The results of ELISA experiment and phage recovery assay provided strong support for the notion that SW-TUP phage binds to polystyrene with a significantly higher affinity than control phage clones. Furthermore, this polystyrene binding was demonstrated to occur in a concentration- and pH-dependent mode. Characterization of the propagation profile of phage clones within a specified time course revealed no statistically significant difference between the amplification rate of SW-TUP and control phages. Our findings lead us to the conclusion that SW-TUP phage clone with the displayed peptide TLHPAAD is not a true target binder and its selection in biopanning experiments results from its bidning affinity to the polystyrene surface of the solid phase.

Evaluation of GalaxyDock Based on the Community Structure-Activity Resource 2013 and 2014 Benchmark Studies.

We analyze the results of the GalaxyDock protein-ligand docking program in the two recent experiments of Community Structure-Activity Resource (CSAR), CSAR 2013 and 2014. GalaxyDock performs global optimization of a modified AutoDock3 energy function by employing the conformational space annealing method. The energy function of GalaxyDock is quite sensitive to atomic clashes. Such energy functions can be effective for sampling physically correct conformations but may not be effective for scoring when conformations are not fully optimized. In phase 1 of CSAR 2013, we successfully selected all four true binders of digoxigenin along with three false positives. However, the energy values were rather high due to insufficient optimization of the conformations docked to homology models. A posteriori relaxation of the model complex structures by GalaxyRefine improved the docking energy values and differentiated the true binders from the false positives better. In the scoring test of CSAR 2013 phase 2, we selected the best poses for each of the two targets. The results of CSAR 2013 phase 3 suggested that an improved method for generating initial conformations for GalaxyDock is necessary for targets involving bulky ligands. Finally, combining existing binding information with GalaxyDock energy-based optimization may be needed for more accurate binding affinity prediction.

Abstract 3654: Targeting binding function-3 site on the androgen receptor to treat Enzalutamide-resistant prostate cancer

Abstract Interest in developing androgen receptor (AR) inhibitors with novel mechanism of action is on the rise since the commercial anti-androgens (including recently approved drug, Enzalutamide) face clinical limitations. The current therapies fail over a period of time because they all target mutation-prone hormone binding pocket on AR to which the receptor has already developed effective resistance mechanisms. Hence, there is a pressing need for novel therapeutics that inhibit the AR through alternative modes of action. Recent studies have identified a novel binding pocket on the surface of AR called binding function-3 (BF3) that is critical for the AR transcriptional activity. In order to develop promising drug-like candidates, we performed structural optimization of our lead AR BF3 inhibitor, followed by a series of experimental validation of the synthetic analogues. As a result, we have discovered a chemical series of quinoline as new lead BF3 inhibitors. One of the most potent inhibitors identified, VPC-13566, demonstrated an IC50 of 0.071μM in AR eGFP transcriptional assay. Confirming it as a true BF3 binder, VPC-13566 displaced Bag-1L peptide, a co-regulator protein that binds to AR BF3 pocket. Additionally, the Biolayer Interferometry assay detected direct reversible interactions between the AR ligand binding domain and the inhibitor. To confirm VPC-13566 binding to the BF3 pocket, a mutagenesis study was performed. Results with mutants F673E, E837A and N833W confirmed that compound did not show any binding to the protein compared with the wild type in a BLI assay. VPC-13566 demonstrated strong anti-proliferative activity against LNCaP and Enzalutamide-resistant prostate cancer cell lines (MR49F) whereas it did not affect the growth of AR independent PC3 cell line. It also reduced prostate specific antigen (PSA) in both LNCaP and MR49F and reduced expression of AR target genes, PSA and TMPRSS2. These findings suggest that VPC-13566 exhibits AR BF3 specific mechanism of action. Furthermore, VPC-13566 significantly reduced AR-dependent growth of tumors and PSA levels in LNCaP and MR49F xenograft models. Based on these outcomes, it can be anticipated that such drug prototypes will lay a foundation for the development of alternative or supplementary small-molecule therapies capable of combating PCa even in its drug resistant forms. Because the emergence of castration resistance is the lethal end stage of the disease, we anticipate that the proposed research will eventually have a substantial impact on patient survival. Citation Format: Ravi Shashi Nayana Munuganti, Mohamed DH Hassona, Eric Leblanc, Emma T. Guns, Paul S. Rennie, Artem Cherkasov. Targeting binding function-3 site on the androgen receptor to treat Enzalutamide-resistant prostate cancer. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 3654. doi:10.1158/1538-7445.AM2015-3654

Comparative assessment of machine-learning scoring functions on PDBbind 2013

Computational docking is the core process of computer-aided drug design (CADD); it aims at predicting the best orientation and conformation of a small molecule (drug ligand) when bound to a target large receptor molecule (protein) in order to form a stable complex molecule. The docking quality is typically measured by a scoring function: a mathematical predictive model that produces a score representing the binding free energy and hence the stability of the resulting complex molecule. An effective scoring function should produce promising drug candidates which can then be synthesized and physically screened using high throughput screening (HTS) process. Therefore, the key to CADD is the design of an efficient highly accurate scoring function. Many traditional techniques have been proposed, however, the performance was generally poor. Only in the last few years the application of the machine learning (ML) technology has been applied in the design of scoring functions; and the results have been very promising.In this paper, we propose 12 scoring functions based on a wide range of ML techniques. We analyze the performance of each on the scoring power (binding affinity prediction), ranking power (relative ranking prediction), docking power (identifying the native binding poses among computer-generated decoys), and screening power (classifying true binders versus negative binders) using the PDBbind 2013 database. We compare our results with the recently published comparative assessment of scoring functions (CASF-2013) of 20 classical scoring functions most of which are implemented in main-stream commercial software. For the scoring and ranking powers, the proposed ML scoring functions depend on a wide range of features (energy terms, pharmacophore, geometrical) that entirely characterize the protein–ligand complexes (about 108 features); these features are extracted from several docking software available in the literature; a feature-space reduction technique, namely, principal component analysis is then applied and the performance is studied accordingly. For the docking and screening powers, the proposed ML scoring functions depend on the geometrical features of the RF-Score (36 features) to utilize a larger number of training complexes (relative to the large number of decoys in the testing set). For the scoring power, the best ML scoring function (RF) achieves a Pearson correlation coefficient between the predicted and experimentally determined binding affinities of 0.704 versus 0.614 achieved by the best classical scoring function (X-ScoreHM). For the ranking power, the best ML scoring function (RF) achieves a Spearman correlation coefficient between the ranks based on the predicted and experimentally determined binding affinities of 0.697 versus 0.626 achieved by the best classical scoring function (X-ScoreHM). For the docking power, the best ML scoring function (BRT) has a success rate in identifying the top best-scored ligand binding pose within 2Å root-mean-square deviation from the native pose of 13.8% versus 81.0% achieved by the best classical scoring function (ChemPLP@GOLD). For the screening power, the best ML scoring function (SVM) has an average enrichment factor and success rate at the top 1% level of 3.76 and 6.45% versus 19.54 and 60% respectively achieved by the best classical scoring function (GlideScore-SP).