Prediction of P-Glycoprotein Substrates by a Support Vector Machine Approach

Feature importance of machine learning prediction models shows structurally active part and important physicochemical features in drug design

The objective of this study was to obtain the indicators of physicochemical parameters and structurally active sites to design new chemical entities with desirable pharmacokinetic profiles by investigating the process by which machine learning prediction models arrive at their decisions, which are called explainable artificial intelligence. First, we developed the prediction models for metabolic stability, CYP inhibition, and P-gp and BCRP substrate recognition using 265 physicochemical parameters for designing the molecular structures. Four important parameters, including the well-known indicator h_logD, are common in some in vitro studies; as such, these can be used to optimize compounds simultaneously to address multiple pharmacokinetic concerns. Next, we developed machine learning models that had been programmed to show structurally active sites. Many types of machine learning models were developed using the results of in vitro metabolic stability study of around 30000 in-house compounds. The metabolic sites of in-house compounds predicted using some prediction models matched experimentally identified metabolically active sites, with a ratio of number of metabolic sites (predicted/actual) of over 90%. These models can be applied to several screening projects. These two approaches can be employed for obtaining lead compounds with desirable pharmacokinetic profiles efficiently.

Multiple machine learning, molecular docking, and ADMET screening approach for identification of selective inhibitors of CYP1B1

Cytochrome P4501B1 is a ubiquitous family protein that is majorly overexpressed in tumors and is responsible for biotransformation-based inactivation of anti-cancer drugs. This inactivation marks the cause of resistance to chemotherapeutics. In the present study, integrated in-silico approaches were utilized to identify selective CYP1B1 inhibitors. To achieve this objective, we initially developed different machine learning models corresponding to two isoforms of the CYP1 family i.e. CYP1A1 and CYP1B1. Subsequently, small molecule databases including ChemBridge, Maybridge, and natural compound library were screened from the selected models of CYP1B1 and CYP1A1. The obtained CYP1B1 inhibitors were further subjected to molecular docking and ADMET analysis. The selectivity of the obtained hits for CYP1B1 over the other isoforms was also judged with molecular docking analysis. Finally, two hits were found to be the most stable which retained key interactions within the active site of CYP1B1 after the molecular dynamics simulations. Novel compound with CYP-D9 and CYP-14 IDs were found to be the most selective CYP1B1 inhibitors which may address the issue of resistance. Moreover, these compounds can be considered as safe agents for further cell-based and animal model studies.Communicated by Ramaswamy H. Sarma.

A Machine Learning-Based Prediction Platform for P-Glycoprotein Modulators and Its Validation by Molecular Docking

P-glycoprotein (P-gp) is an important determinant of multidrug resistance (MDR) because its overexpression is associated with increased efflux of various established chemotherapy drugs in many clinically resistant and refractory tumors. This leads to insufficient therapeutic targeting of tumor populations, representing a major drawback of cancer chemotherapy. Therefore, P-gp is a target for pharmacological inhibitors to overcome MDR. In the present study, we utilized machine learning strategies to establish a model for P-gp modulators to predict whether a given compound would behave as substrate or inhibitor of P-gp. Random forest feature selection algorithm-based leave-one-out random sampling was used. Testing the model with an external validation set revealed high performance scores. A P-gp modulator list of compounds from the ChEMBL database was used to test the performance, and predictions from both substrate and inhibitor classes were selected for the last step of validation with molecular docking. Predicted substrates revealed similar docking poses than that of doxorubicin, and predicted inhibitors revealed similar docking poses than that of the known P-gp inhibitor elacridar, implying the validity of the predictions. We conclude that the machine-learning approach introduced in this investigation may serve as a tool for the rapid detection of P-gp substrates and inhibitors in large chemical libraries.

Computational approaches in cancer multidrug resistance research: Identification of potential biomarkers, drug targets and drug-target interactions

Like physics in the 19th century, biology and molecular biology in particular, has been fertilized and enhanced like few other scientific fields, by the incorporation of mathematical methods. In the last decades, a whole new scientific field, bioinformatics, has developed with an output of over 30,000 papers a year (Pubmed search using the keyword "bioinformatics"). Huge databases of mass throughput data have been established, with ArrayExpress alone containing more than 2.7 million assays (October 2019). Computational methods have become indispensable tools in molecular biology, particularly in one of the most challenging areas of cancer research, multidrug resistance (MDR). However, confronted with a plethora of different algorithms, approaches, and methods, the average researcher faces key questions: Which methods do exist? Which methods can be used to tackle the aims of a given study? Or, more generally, how do I use computational biology/bioinformatics to bolster my research? The current review is aimed at providing guidance to existing methods with relevance to MDR research. In particular, we provide an overview on: a) the identification of potential biomarkers using expression data; b) the prediction of treatment response by machine learning methods; c) the employment of network approaches to identify gene/protein regulatory networks and potential key players; d) the identification of drug-target interactions; e) the use of bipartite networks to identify multidrug targets; f) the identification of cellular subpopulations with the MDR phenotype; and, finally, g) the use of molecular modeling methods to guide and enhance drug discovery. This review shall serve as a guide through some of the basic concepts useful in MDR research. It shall give the reader some ideas about the possibilities in MDR research by using computational tools, and, finally, it shall provide a short overview of relevant literature.

Machine Learning Models Combined with Virtual Screening and Molecular Docking to Predict Human Topoisomerase I Inhibitors

In this work, random forest (RF), support vector machine, k-nearest neighbor and C4.5 decision tree, were used to establish classification models for predicting whether an unknown molecule is an inhibitor of human topoisomerase I (Top1) protein. All these models have achieved satisfactory results, with total prediction accuracies from 89.70% to 97.12%. Through comparative analysis, it can be found that the RF model has the best forecasting effect. The parameters were further optimized to generate the best-performing RF model. At the same time, features selection was implemented to choose properties most relevant to the inhibition of Top1 from 189 molecular descriptors through a special RF procedure. Subsequently, a ligand-based virtual screening was performed from the Maybridge database by the optimal RF model and 596 hits were picked out. Then, 67 molecules with relative probability scores over 0.7 were selected based on the screening results. Next, the 67 molecules above were docked to Top1 using AutoDock Vina. Finally, six top-ranked molecules with binding energies less than −10.0 kcal/mol were screened out and a common backbone, which is entirely different from that of existing Top1 inhibitors reported in the literature, was found.

Development of Simplified in Vitro P-Glycoprotein Substrate Assay and in Silico Prediction Models To Evaluate Transport Potential of P-Glycoprotein

For efficient drug discovery and screening, it is necessary to simplify P-glycoprotein (P-gp) substrate assays and to provide in silico models that predict the transport potential of P-gp. In this study, we developed a simplified in vitro screening method to evaluate P-gp substrates by unidirectional membrane transport in P-gp-overexpressing cells. The unidirectional flux ratio positively correlated with parameters of the conventional bidirectional P-gp substrate assay ( R² = 0.941) and in vivo K_p,brain ratio (mdr1a/1b KO/WT) in mice ( R² = 0.800). Our in vitro P-gp substrate assay had high reproducibility and required approximately half the labor of the conventional method. We also constructed regression models to predict the value of P-gp-mediated flux and three-class classification models to predict P-gp substrate potential (low-, medium-, and high-potential) using 2397 data entries with the largest data set collected under the same experimental conditions. Most compounds in the test set fell within two- and three-fold errors in the random forest regression model (71.3 and 88.5%, respectively). Furthermore, the random forest three-class classification model showed a high balanced accuracy of 0.821 and precision of 0.761 for the low-potential classes in the test set. We concluded that the simplified in vitro P-gp substrate assay was suitable for compound screening in the early stages of drug discovery and that the in silico regression model and three-class classification model using only chemical structure information could identify the transport potential of compounds including P-gp-mediated flux ratios. Our proposed method is expected to be a practical tool to optimize effective central nervous system (CNS) drugs, to avoid CNS side effects, and to improve intestinal absorption.

Boosted feature selectors: a case study on prediction P-gp inhibitors and substrates

Feature selection is commonly used as a preprocessing step to machine learning for improving learning performance, lowering computational complexity and facilitating model interpretation. This paper proposes the application of boosting feature selection to improve the classification performance of standard feature selection algorithms evaluated for the prediction of P-gp inhibitors and substrates. Two well-known classification algorithms, decision trees and support vector machines, were used to classify the chemical compounds. The experimental results showed better performance for boosting feature selection with respect to the standard feature selection algorithms while maintaining the capability for feature reduction.

Theoretical Prediction of the Complex P-Glycoprotein Substrate Efflux Based on the Novel Hierarchical Support Vector Regression Scheme

P-glycoprotein (P-gp), a membrane-bound transporter, can eliminate xenobiotics by transporting them out of the cells or blood⁻brain barrier (BBB) at the expense of ATP hydrolysis. Thus, P-gp mediated efflux plays a pivotal role in altering the absorption and disposition of a wide range of substrates. Nevertheless, the mechanism of P-gp substrate efflux is rather complex since it can take place through active transport and passive permeability in addition to multiple P-gp substrate binding sites. A nonlinear quantitative structure⁻activity relationship (QSAR) model was developed in this study using the novel machine learning-based hierarchical support vector regression (HSVR) scheme to explore the perplexing relationships between descriptors and efflux ratio. The predictions by HSVR were found to be in good agreement with the observed values for the molecules in the training set (n = 50, r² = 0.96, qCV2 = 0.94, RMSE = 0.10, s = 0.10) and test set (n = 13, q² = 0.80⁻0.87, RMSE = 0.21, s = 0.22). When subjected to a variety of statistical validations, the developed HSVR model consistently met the most stringent criteria. A mock test also asserted the predictivity of HSVR. Consequently, this HSVR model can be adopted to facilitate drug discovery and development.

A novel adaptive ensemble classification framework for ADME prediction

AECF is a GA based ensemble method. It includes four components which are (1) data balancing, (2) generating individual models, (3) combining individual models, and (4) optimizing the ensemble.

Discovery of novel dual VEGFR2 and Src inhibitors using a multistep virtual screening approach

AIM

Simultaneous inhibition of VEGFR2 and Src may enhance the efficacy of VEGFR2-targeted cancer therapeutics. Hence, development of dual inhibitors on VEGFR2 and Src can be a useful strategy for such treatments.

MATERIALS & METHODS

A multistep virtual screening protocol, comprising ligand-based support vector machines method, drug-likeness rules filter and structure-based molecular docking, was developed and employed to identify dual inhibitors of VEGFR2 and Src from a large commercial chemical library. Kinase inhibitory assays and cell viability assays were then used for experimental validation.

RESULTS

A set of compounds belonging to six different molecular scaffolds was identified and sent for biological evaluation. Compound 3c belonging to the 2-amino-3-cyanopyridine scaffold exhibited good antiproliferative effect and dual-target activities against VEGFR2 and Src.

CONCLUSION

This study demonstrated the ability of the multistep virtual screening approach to identify novel multitarget agents.

The Next Era: Deep Learning in Pharmaceutical Research

Over the past decade we have witnessed the increasing sophistication of machine learning algorithms applied in daily use from internet searches, voice recognition, social network software to machine vision software in cameras, phones, robots and self-driving cars. Pharmaceutical research has also seen its fair share of machine learning developments. For example, applying such methods to mine the growing datasets that are created in drug discovery not only enables us to learn from the past but to predict a molecule's properties and behavior in future. The latest machine learning algorithm garnering significant attention is deep learning, which is an artificial neural network with multiple hidden layers. Publications over the last 3 years suggest that this algorithm may have advantages over previous machine learning methods and offer a slight but discernable edge in predictive performance. The time has come for a balanced review of this technique but also to apply machine learning methods such as deep learning across a wider array of endpoints relevant to pharmaceutical research for which the datasets are growing such as physicochemical property prediction, formulation prediction, absorption, distribution, metabolism, excretion and toxicity (ADME/Tox), target prediction and skin permeation, etc. We also show that there are many potential applications of deep learning beyond cheminformatics. It will be important to perform prospective testing (which has been carried out rarely to date) in order to convince skeptics that there will be benefits from investing in this technique.

Development of a Support Vector Machine-Based System to Predict Whether a Compound Is a Substrate of a Given Drug Transporter Using Its Chemical Structure

The aim of this study was to develop an in silico prediction system to assess which of 7 categories of drug transporters (organic anion transporting polypeptide [OATP] 1B1/1B3, multidrug resistance-associated protein [MRP] 2/3/4, organic anion transporter [OAT] 1, OAT3, organic cation transporter [OCT] 1/2/multidrug and toxin extrusion [MATE] 1/2-K, multidrug resistance protein 1 [MDR1], and breast cancer resistance protein [BCRP]) can recognize compounds as substrates using its chemical structure alone. We compiled an internal data set consisting of 260 compounds that are substrates for at least 1 of the 7 categories of drug transporters. Four physicochemical parameters (charge, molecular weight, lipophilicity, and plasma unbound fraction) of each compound were used as the basic descriptors. Furthermore, a greedy algorithm was used to select 3 additional physicochemical descriptors from 731 available descriptors. In addition, transporter nonsubstrates tend not to be in the public domain; we, thus, tried to compile an expert-curated data set of putative nonsubstrates for each transporter using personal opinions of 11 researchers in the field of drug transporters. The best prediction was finally achieved by a support vector machine based on 4 basic and 3 additional descriptors. The model correctly judged that 364 of 412 compounds (internal data set) and 111 of 136 compounds (external data set) were substrates, indicating that this model performs well enough to predict the specificity of transporter substrates.

Cell permeability beyond the rule of 5

Drug discovery for difficult targets that have large and flat binding sites is often better suited to compounds beyond the "rule of 5" (bRo5). However, such compounds carry higher pharmacokinetic risks, such as low solubility and permeability, and increased efflux and metabolism. Interestingly, recent drug approvals and studies suggest that cell permeable and orally bioavailable drugs can be discovered far into bRo5 space. Tactics such as reduction or shielding of polarity by N-methylation, bulky side chains and intramolecular hydrogen bonds may be used to increase cell permeability in this space, but often results in decreased solubility. Conformationally flexible compounds can, however, combine high permeability and solubility, properties that are keys for cell permeability and intestinal absorption. Recent developments in computational conformational analysis will aid design of such compounds and hence prediction of cell permeability. Transporter mediated efflux occurs for most investigated drugs in bRo5 space, however it is commonly overcome by high local intestinal concentrations on oral administration. In contrast, there is little data to support significant impact of transporter-mediated intestinal absorption in bRo5 space. Current knowledge of compound properties that govern transporter effects of bRo5 drugs is limited and requires further fundamental and comprehensive studies.

Recent progresses in the exploration of machine learning methods as in-silico ADME prediction tools

In-silico methods have been explored as potential tools for assessing ADME and ADME regulatory properties particularly in early drug discovery stages. Machine learning methods, with their ability in classifying diverse structures and complex mechanisms, are well suited for predicting ADME and ADME regulatory properties. Recent efforts have been directed at the broadening of application scopes and the improvement of predictive performance with particular focuses on the coverage of ADME properties, and exploration of more diversified training data, appropriate molecular features, and consensus modeling. Moreover, several online machine learning ADME prediction servers have emerged. Here we review these progresses and discuss the performances, application prospects and challenges of exploring machine learning methods as useful tools in predicting ADME and ADME regulatory properties.

Homology modelling and molecular docking of MDR1 with chemotherapeutic agents in non-small cell lung cancer

MDR1, a protein commonly involved in drug transport, has been linked to multi drug resistance and disease progression in cancers such as non-small cell lung cancer. Hence, targeting this protein is essential for improving drug design and preventing adverse drug-drug interactions. The aim of the study was to examine chemotherapeutic drug binding to MDR1 and the interactions therein. We have used Schrödinger suite 2014, to perform homology modelling of human MDR1 based on Mouse MDR1, followed by Induced Fit Docking with Paclitaxel, Docetaxel, Gemcitabine, Carboplatin and Cisplatin drugs. Finally, we evaluated drug binding affinities using Prime/MMGBSA and using these scores we compared the affinities of combination therapies against MDR1. Analysis of the docking results showed Paclitaxel>Docetaxel>Gemcitabine>Carboplatin>Cisplatin as the order of binding affinities, with Paclitaxel having the best docking score. The combination drug binding affinity analysis showed Paclitaxel+Gemcitabine to have the best docking score and hence, efficacy. Through our investigation we have identified the residues Gln 195 and Gln 946 to be more frequently involved in drug binding interactions with MDR1. Our results suggest that, Paclitaxel or combination of Paclitaxel+Gemcitabine could serve as a suitable therapy against MDR1 in NSCLC patients. Thus, our study provides new insight into the possible repurposing of chemotherapeutic drugs in targeting elevated MDR1 levels in NSCLC patients, thereby ensuring better overall outcome. Further our study highlights the use of in silico methodologies in understanding drug binding to protein targets and its relevance to advancing lung cancer therapy.

Three- and four-class classification models for P-glycoprotein inhibitors using counter-propagation neural networks

P-glycoprotein (P-gp) is an ATP binding cassette (ABC) transporter that helps to protect several certain human organs from xenobiotic exposure. This efflux pump is also responsible for multi-drug resistance (MDR), an issue of the chemotherapy approach in the fight against cancer. Therefore, the discovery of P-gp inhibitors is considered one of the most popular strategies to reverse MDR in tumour cells and to improve therapeutic efficacy of commonly used cytotoxic drugs. Until now, several generations of P-gp inhibitors have been developed but they have largely failed in preclinical and clinical studies due to lack of selectivity, poor solubility and severe pharmacokinetic interactions. In this study, three models (SION, SIO, SIN) to classify specific 'true' P-gp inhibitors as well as three other models (CPBN, CPB1, CPN) to distinguish between P-gp inhibitors, CYP 3A inhibitors and co-inhibitors of these proteins with rather high accuracy values for the test set and the external set were generated based on counter-propagation neural networks (CPG-NN). Such three and four-class classification models helped provide more information about the bioactivities of compounds not only on one target (P-gp), but also on a combination of multiple targets (P-gp, CYP 3A).

Multivariate statistical analysis methods in QSAR

The emphasis of this review is particularly on multivariate statistical methods currently used in quantitative structure–activity relationship (QSAR) studies.

Computational classification models for predicting the interaction of drugs with P-glycoprotein and breast cancer resistance protein

P-glycoprotein (P-gp/ABCB1) and breast cancer resistance protein (BCRP/ABCG2) are two members of the adenosine triphosphate (ATP) binding cassette (ABC) family of transporters which function as membrane efflux transporters and display considerable substrate promiscuity. Both are known to significantly influence the absorption, distribution and elimination of drugs, mediate drug-drug interactions and contribute to multiple drug resistance (MDR) of cancer cells. Correspondingly, timely characterization of the interaction of novel leads and drug candidates with these two transporters is of great importance. In this study, several computational classification models for prediction of transport and inhibition of P-gp and BCRP, respectively, were developed based on newly compiled and critically evaluated experimental data. Artificial neural network (ANN) and support vector machine (SVM) ensemble based models were explored, as well as knowledge-based approaches to descriptor selection. The average overall classification accuracy of best performing models was 82% for P-gp transport, 88% for BCRP transport, 89% for P-gp inhibition and 87% for BCRP inhibition, determined across an array of different test sets. An analysis of substrate overlap between P-gp and BCRP was also performed. The accuracy, simplicity and interpretability of the proposed models suggest that they could be of significant utility in the drug discovery and development settings.

A prediction model of drug-induced ototoxicity developed by an optimal support vector machine (SVM) method

Drug-induced ototoxicity, as a toxic side effect, is an important issue needed to be considered in drug discovery. Nevertheless, current experimental methods used to evaluate drug-induced ototoxicity are often time-consuming and expensive, indicating that they are not suitable for a large-scale evaluation of drug-induced ototoxicity in the early stage of drug discovery. We thus, in this investigation, established an effective computational prediction model of drug-induced ototoxicity using an optimal support vector machine (SVM) method, GA-CG-SVM. Three GA-CG-SVM models were developed based on three training sets containing agents bearing different risk levels of drug-induced ototoxicity. For comparison, models based on naïve Bayesian (NB) and recursive partitioning (RP) methods were also used on the same training sets. Among all the prediction models, the GA-CG-SVM model II showed the best performance, which offered prediction accuracies of 85.33% and 83.05% for two independent test sets, respectively. Overall, the good performance of the GA-CG-SVM model II indicates that it could be used for the prediction of drug-induced ototoxicity in the early stage of drug discovery.

Accurate models for P-gp drug recognition induced from a cancer cell line cytotoxicity screen

P-glycoprotein (P-gp, MDR1) is a promiscuous drug efflux pump of substantial pharmacological importance. Taking advantage of large-scale cytotoxicity screening data involving 60 cancer cell lines, we correlated the differential biological activities of ∼13,000 compounds against cellular P-gp levels. We created a large set of 934 high-confidence P-gp substrates or nonsubstrates by enforcing agreement with an orthogonal criterion involving P-gp overexpressing ADR-RES cells. A support vector machine (SVM) was 86.7% accurate in discriminating P-gp substrates on independent test data, exceeding previous models. Two molecular features had an overarching influence: nearly all P-gp substrates were large (>35 atoms including H) and dense (specific volume of <7.3 Å(3)/atom) molecules. Seven other descriptors and 24 molecular fragments ("effluxophores") were found enriched in the (non)substrates and incorporated into interpretable rule-based models. Biological experiments on an independent P-gp overexpressing cell line, the vincristine-resistant VK2, allowed us to reclassify six compounds previously annotated as substrates, validating our method's predictive ability. Models are freely available at http://pgp.biozyne.com .

Predicting substrates of the human breast cancer resistance protein using a support vector machine method

Background

Human breast cancer resistance protein (BCRP) is an ATP-binding cassette (ABC) efflux transporter that confers multidrug resistance in cancers and also plays an important role in the absorption, distribution and elimination of drugs. Prediction as to if drugs or new molecular entities are BCRP substrates should afford a cost-effective means that can help evaluate the pharmacokinetic properties, efficacy, and safety of these drugs or drug candidates. At present, limited studies have been done to develop in silico prediction models for BCRP substrates.

In this study, we developed support vector machine (SVM) models to predict wild-type BCRP substrates based on a total of 263 known BCRP substrates and non-substrates collected from literature. The final SVM model was integrated to a free web server.

Results

We showed that the final SVM model had an overall prediction accuracy of ~73% for an independent external validation data set of 40 compounds. The prediction accuracy for wild-type BCRP substrates was ~76%, which is higher than that for non-substrates. The free web server (http://bcrp.althotas.com) allows the users to predict whether a query compound is a wild-type BCRP substrate and calculate its physicochemical properties such as molecular weight, logP value, and polarizability.

Conclusions

We have developed an SVM prediction model for wild-type BCRP substrates based on a relatively large number of known wild-type BCRP substrates and non-substrates. This model may prove valuable for screening substrates and non-substrates of BCRP, a clinically important ABC efflux drug transporter.

JFeature

Prediction of various functional properties of proteins has long been a central theme of bioinformatics in the post-genomic era. Statistical learning, in addition to analysis based on sequence similarity, was proven successful to detect complex sequence-function associations in many applications. JFeature is an integrated Java tool to facilitate extraction of global sequence features and preparation of example sets, in statistical learning studies of sequence-function relationships. With a user-friendly graphical interface, it computes the composition, distribution, transition and auto-correlation features from sequence. It also helps to assemble a negative example set based on the most-dissimilar principle. The Java package and supplementary documentations are available at http://www.cls.zju.edu.cn/rlibs/software/jfeature.html.

QSAR and Predictors of Eye and Skin Effects

In this study, the ensemble of features and training samples was examined with a collection of support vector machines. The effects of data sampling methods, ratio of positive to negative compounds, and types of base models combiner to produce ensemble models were explored. The ensemble method was applied to produce four separate in silico models to classify the labels for eye/skin corrosion (H314), skin irritation (H315), serious eye damage (H318), and eye irritation (H319), which are defined in the "Globally Harmonized System of Classification and Labelling of Chemicals". To the best of our knowledge, the training set used in this work is one of the largest (made of publicly available data) with acceptable prediction performances. These models were distributed via PaDEL-DDPredictor (http://padel.nus.edu.sg/software/padelddpredictor) that can be downloaded freely for public use.

Development and experimental test of support vector machines virtual screening method for searching Src inhibitors from large compound libraries

BACKGROUND

Src plays various roles in tumour progression, invasion, metastasis, angiogenesis and survival. It is one of the multiple targets of multi-target kinase inhibitors in clinical uses and trials for the treatment of leukemia and other cancers. These successes and appearances of drug resistance in some patients have raised significant interest and efforts in discovering new Src inhibitors. Various in-silico methods have been used in some of these efforts. It is desirable to explore additional in-silico methods, particularly those capable of searching large compound libraries at high yields and reduced false-hit rates.

RESULTS

We evaluated support vector machines (SVM) as virtual screening tools for searching Src inhibitors from large compound libraries. SVM trained and tested by 1,703 inhibitors and 63,318 putative non-inhibitors correctly identified 93.53%~ 95.01% inhibitors and 99.81%~ 99.90% non-inhibitors in 5-fold cross validation studies. SVM trained by 1,703 inhibitors reported before 2011 and 63,318 putative non-inhibitors correctly identified 70.45% of the 44 inhibitors reported since 2011, and predicted as inhibitors 44,843 (0.33%) of 13.56M PubChem, 1,496 (0.89%) of 168 K MDDR, and 719 (7.73%) of 9,305 MDDR compounds similar to the known inhibitors.

CONCLUSIONS

SVM showed comparable yield and reduced false hit rates in searching large compound libraries compared to the similarity-based and other machine-learning VS methods developed from the same set of training compounds and molecular descriptors. We tested three virtual hits of the same novel scaffold from in-house chemical libraries not reported as Src inhibitor, one of which showed moderate activity. SVM may be potentially explored for searching Src inhibitors from large compound libraries at low false-hit rates.

Fingerprint-based in silico models for the prediction of P-glycoprotein substrates and inhibitors

P-Glycoprotein (P-gp, ABCB1) plays a significant role in determining the ADMET properties of drugs and drug candidates. Substrates of P-gp are not only subject to multidrug resistance (MDR) in tumor therapy, they are also associated with poor pharmacokinetic profiles. In contrast, inhibitors of P-gp have been advocated as modulators of MDR. However, due to the polyspecificity of P-gp, knowledge on the molecular basis of ligand-transporter interaction is still poor, which renders the prediction of whether a compound is a P-gp substrate/non-substrate or an inhibitor/non-inhibitor quite challenging. In the present investigation, we used a set of fingerprints representing the presence/absence of various functional groups for machine learning based classification of a set of 484 substrates/non-substrates and a set of 1935 inhibitors/non-inhibitors. Best models were obtained using a combination of a wrapper subset evaluator (WSE) with random forest (RF), kappa nearest neighbor (kNN) and support vector machine (SVM), showing accuracies >70%. Best P-gp substrate models were further validated with three sets of external P-gp substrate sources, which include Drug Bank (n = 134), TP Search (n = 90) and a set compiled from literature (n = 76). Association rule analysis explores the various structural feature requirements for P-gp substrates and inhibitors.

Qualitative prediction of blood-brain barrier permeability on a large and refined dataset

The prediction of blood-brain barrier permeation is vitally important for the optimization of drugs targeting the central nervous system as well as for avoiding side effects of peripheral drugs. Following a previously proposed model on blood-brain barrier penetration, we calculated the cross-sectional area perpendicular to the amphiphilic axis. We obtained a high correlation between calculated and experimental cross-sectional area (r = 0.898, n = 32). Based on these results, we examined a correlation of the calculated cross-sectional area with blood-brain barrier penetration given by logBB values. We combined various literature data sets to form a large-scale logBB dataset with 362 experimental logBB values. Quantitative models were calculated using bootstrap validated multiple linear regression. Qualitative models were built by a bootstrapped random forest algorithm. Both methods found similar descriptors such as polar surface area, pKa, logP, charges and number of positive ionisable groups to be predictive for logBB. In contrast to our initial assumption, we were not able to obtain models with the cross-sectional area chosen as relevant parameter for both approaches. Comparing those two different techniques, qualitative random forest models are better suited for blood-brain barrier permeability prediction, especially when reducing the number of descriptors and using a large dataset. A random forest prediction system (n(trees) = 5) based on only four descriptors yields a validated accuracy of 88%.

Predicting P-glycoprotein-mediated drug transport based on support vector machine and three-dimensional crystal structure of P-glycoprotein

Human P-glycoprotein (P-gp) is an ATP-binding cassette multidrug transporter that confers resistance to a wide range of chemotherapeutic agents in cancer cells by active efflux of the drugs from cells. P-gp also plays a key role in limiting oral absorption and brain penetration and in facilitating biliary and renal elimination of structurally diverse drugs. Thus, identification of drugs or new molecular entities to be P-gp substrates is of vital importance for predicting the pharmacokinetics, efficacy, safety, or tissue levels of drugs or drug candidates. At present, publicly available, reliable in silico models predicting P-gp substrates are scarce. In this study, a support vector machine (SVM) method was developed to predict P-gp substrates and P-gp-substrate interactions, based on a training data set of 197 known P-gp substrates and non-substrates collected from the literature. We showed that the SVM method had a prediction accuracy of approximately 80% on an independent external validation data set of 32 compounds. A homology model of human P-gp based on the X-ray structure of mouse P-gp as a template has been constructed. We showed that molecular docking to the P-gp structures successfully predicted the geometry of P-gp-ligand complexes. Our SVM prediction and the molecular docking methods have been integrated into a free web server (http://pgp.althotas.com), which allows the users to predict whether a given compound is a P-gp substrate and how it binds to and interacts with P-gp. Utilization of such a web server may prove valuable for both rational drug design and screening.

A prediction model of substrates and non-substrates of breast cancer resistance protein (BCRP) developed by GA-CG-SVM method

Breast cancer resistance protein (BCRP) is one of the key multi-drug resistance proteins, which significantly influences the therapeutic effects of many drugs, particularly anti-cancer drugs. Thus, distinguishing between substrates and non-substrates of BCRP is important not only for clinical use but also for drug discovery and development. In this study, a prediction model of the substrates and non-substrates of BCRP was developed using a modified support vector machine (SVM) method, namely GA-CG-SVM. The overall prediction accuracy of the established GA-CG-SVM model is 91.3% for the training set and 85.0% for an independent validation set. For comparison, two other machine learning methods, namely, C4.5 DT and k-NN, were also adopted to build prediction models. The results show that the GA-CG-SVM model is significantly superior to C4.5 DT and k-NN models in terms of the prediction accuracy. To sum up, the prediction model of BCRP substrates and non-substrates generated by the GA-CG-SVM method is sufficiently good and could be used as a screening tool for identifying the substrates and non-substrates of BCRP.

Using kernel alignment to select features of molecular descriptors in a QSAR study

Quantitative structure-activity relationships (QSARs) correlate biological activities of chemical compounds with their physicochemical descriptors. By modeling the observed relationship seen between molecular descriptors and their corresponding biological activities, we may predict the behavior of other molecules with similar descriptors. In QSAR studies, it has been shown that the quality of the prediction model strongly depends on the selected features within molecular descriptors. Thus, methods capable of automatic selection of relevant features are very desirable. In this paper, we present a new feature selection algorithm for a QSAR study based on kernel alignment which has been used as a measure of similarity between two kernel functions. In our algorithm, we deploy kernel alignment as an evaluation tool, using recursive feature elimination to compute a molecular descriptor containing the most important features needed for a classification application. Empirical results show that the algorithm works well for the computation of descriptors for various applications involving different QSAR data sets. The prediction accuracies are substantially increased and are comparable to those from earlier studies.

Use of shape similarities for the classification of P-glycoprotein substrates and nonsubstrates

Background: The multidrug transporter P-glycoprotein (P-gp) ATP-binding cassette B1 (ABCB1) is one of the key proteins influencing bioavailability and uptake of drugs in the brain. In addition, it is one of the main factors contributing to multidrug resistance in tumor therapy. Due to its promiscuous substrate recognition, prediction of substrate properties for the multidrug transporter P-gp represents a challenging task. Results: Here, we present data on three classification methods of ABCB1 substrates and nonsubstrates based on 2D and 3D shape similarity calculations with special emphasis on the use of the similarity-based relationship approach. The results indicate that a reference set structurally similar to the data set performs superiorly to those selected on the basis of maximum diversity and suggests Random Forest as the most suitable classification method for this data set. Conclusion: This study suggests 2D descriptors representing 3D features best suited to the classification of P-gp substrates and nonsubstrates.

P-glycoprotein substrate models using support vector machines based on a comprehensive data set

P-glycoprotein (P-gp) is one of the major ABC transporters and involved in many essential processes such as lipid and steroid transport across cell membranes but also in the uptake of drugs such as HIV protease and reverse transcriptase inhibitors. Despite its importance, reliable models predicting substrates of P-gp are scarce. In this study, we have built several computational models to predict whether or not a compound is a P-gp substrate, based on the largest data set yet published, employing 332 distinct structures. Each molecule is represented by ADRIANA.Code, MOE, and ECFP_4 fingerprint descriptors. The models are computed using a support vector machine based on a training set which includes 131 substrates and 81 nonsubstrates that were evaluated by 5-, 10-fold, and leave-one-out (LOO) cross-validation. The best model gives a Matthews Correlation Coefficient of 0.73 and a prediction accuracy of 0.88 on the test set. Examination of the model based on ECFP_4 fingerprints revealed several substructures which could have significance in separating substrates and nonsubstrates of P-gp, such as the nitrile and sulfoxide functional groups which have a higher frequency in nonsubstrates than in substrates. In addition structural isomerism in sugars was found to result in remarkable differences regarding the likelihood of a compound to be a substrate for P-gp.

Amino Acid Features of P1B-ATPase Heavy Metal Transporters Enabling Small Numbers of Organisms to Cope with Heavy Metal Pollution

Phytoremediation refers to the use of plants for extraction and detoxification of pollutants, providing a new and powerful weapon against a polluted environment. In some plants, such as Thlaspi spp, heavy metal ATPases are involved in overall metal ion homeostasis and hyperaccumulation. P1B-ATPases pump a wide range of cations, especially heavy metals, across membranes against their electrochemical gradients. Determination of the protein characteristics of P1B-ATPases in hyperaccumulator plants provides a new opportuntity for engineering of phytoremediating plants. In this study, using diverse weighting and modeling approaches, 2644 protein characteristics of primary, secondary, and tertiary structures of P1B-ATPases in hyperaccumulator and nonhyperaccumulator plants were extracted and compared to identify differences between proteins in hyperaccumulator and nonhyperaccumulator pumps. Although the protein characteristics were variable in their weighting, tree and rule induction models; glycine count, frequency of glutamine-valine, and valine-phenylalanine count were the most important attributes highlighted by 10, five, and four models, respectively. In addition, a precise model was built to discriminate P1B-ATPases in different organisms based on their structural protein features. Moreover, reliable models for prediction of the hyperaccumulating activity of unknown P1B-ATPase pumps were developed. Uncovering important structural features of hyperaccumulator pumps in this study has provided the knowledge required for future modification and engineering of these pumps by techniques such as site-directed mutagenesis.

Diverse approaches for the enhancement of oral drug bioavailability

In conscious and co-operating patients, oral drug delivery remains the preferable route of drug administration. However, not all drugs possess the desirable physicochemical and pharmacokinetic properties which favor oral administration mainly due to poor bioavailability. This has in some cases led to the choice of other routes of administration, which may compromise the convenience and increase the risk of non-compliance. Poor bioavailability has necessitated the administration of higher than normally required oral doses which often leads to economic wastages, risk of toxicity, erratic and unpredictable responses. The challenge over the years has been to design techniques that will allow oral administration of most drugs, irrespective of their properties, to achieve a therapeutic systemic availability. This will be a worthy achievement since over 90% of therapeutic compounds are known to possess oral bioavailability limitations. In this review, an attempt has been made to explore various approaches that have been used in recent years to improve oral drug bioavailability, including physical and chemical means. This review strives to provide a comprehensive overview of advances made over the past 10 years (2000-2010) in the improvement of the oral bioavailability of drugs. Briefly, the design of prodrugs to bypass metabolism or to enhance solubility as well as modification of formulation techniques such as the use of additives, permeation enhancers, solubilizers, emulsifiers and non-aqueous vehicles have been discussed. Arising approaches, such as formulation modification techniques; novel drug delivery systems, which exploit the gastrointestinal regionality of drugs, and include the pharmaceutical application of nanotechnology as an emerging area in drug delivery; inhibition of efflux pumps; and inhibition of presystemic metabolism have been more extensively addressed. This critical review sought to assess each method aimed at enhancing the oral bioavailability of drugs in terms of the purpose, scientific basis, limitations, commercial application, as well as the areas in which current research efforts are being focused and should be focused in the future.

Homology modeling and binding site assessment of the human P-glycoprotein

Background: Due to its impact on multidrug resistance and pharmacokinetics P-glycoprotein (P-gp) has been identified as an important anti-target in pharmaceutical research. Recent publication of the mouse P-gp structure prompted us to build a new model for human P-gp and investigate its binding-site characteristics. Results: We developed and validated the human P-gp model that was used for induced-fit docking of experimentally characterized P-gp substrates. Residues located in the binding pocket are in good correlation with the results of side-directed mutagenesis studies. However, enrichment studies aimed at discriminating inhibitors and substrates from decoys resulted in only limited enrichments. Conclusion: A mouse P-gp-based homology model might be useful when analyzing protein–ligand interactions of known human P-gp substrates if induced-fit effects are considered.