Relevant Descriptors Research Articles

Prediction of therapeutic potency of tacrine derivatives as BuChE inhibitors from quantitative structure–activity relationship modelling

Numerous studies show that tacrine derivatives exhibit increased inhibitory activity against butyrylcholinesterase (BuChE) and acetylcholinesterase (AChE). However, the screening assays for currently available BuChE inhibitors are expensive, time consuming and dependent on the inhibitory compound. It is therefore desirable to develop alternative methods to facilitate the screening of these derivatives in the early phase of drug discovery. In order to develop robust predictive models, three regression methods were chosen in this study: multiple linear regression (MLR), support vector regression (SVR) and multilayer perceptron network (MLP). Eight relevant descriptors were selected on a dataset of 151 molecules using a method based on genetic algorithms. Internal and external validation strategies play an important role. Also, to check the robustness of the selected models, all available validation strategies were used, and all criteria used to validate these models revealed the superiority of the SVR model. The statistical parameters obtained with the SVR model were RMSE = 0.197, r2 = 0.969 and Q2 = 0.964 for the training set, and r2 = 0.906 and Q2 = 0.891 for the test set. Therefore, the model developed in this study provides an excellent prediction of the inhibitory concentration of tacrine derivatives.

Quantitative Structure-Activity Relationship Model for HCVNS5B inhibitors based on an Antlion Optimizer-Adaptive Neuro-Fuzzy Inference System

The global prevalence of hepatitis C Virus (HCV) is approximately 3% and one-fifth of all HCV carriers live in the Middle East, where Egypt has the highest global incidence of HCV infection. Quantitative structure-activity relationship (QSAR) models were used in many applications for predicting the potential effects of chemicals on human health and environment. The adaptive neuro-fuzzy inference system (ANFIS) is one of the most popular regression methods for building a nonlinear QSAR model. However, the quality of ANFIS is influenced by the size of the descriptors, so descriptor selection methods have been proposed, although these methods are affected by slow convergence and high time complexity. To avoid these limitations, the antlion optimizer was used to select relevant descriptors, before constructing a nonlinear QSAR model based on the PIC50 and these descriptors using ANFIS. In our experiments, 1029 compounds were used, which comprised 579 HCVNS5B inhibitors (PIC50 < ~14) and 450 non-HCVNS5B inhibitors (PIC50 > ~14). The experimental results showed that the proposed QSAR model obtained acceptable accuracy according to different measures, where {{boldsymbol{R}}}^{{boldsymbol{2}}} was 0.952 and 0.923 for the training and testing sets, respectively, using cross-validation, while {{boldsymbol{R}}}_{{boldsymbol{L}}{boldsymbol{O}}{boldsymbol{O}}}^{{boldsymbol{2}}} was 0.8822 using leave-one-out (LOO).

Promises of Machine Learning Approaches in Prediction of Absorption of Compounds.

The Machine Learning (ML) is one of the fastest developing techniques in the prediction and evaluation of important pharmacokinetic properties such as absorption, distribution, metabolism and excretion. The availability of a large number of robust validation techniques for prediction models devoted to pharmacokinetics has significantly enhanced the trust and authenticity in ML approaches. There is a series of prediction models generated and used for rapid screening of compounds on the basis of absorption in last one decade. Prediction of absorption of compounds using ML models has great potential across the pharmaceutical industry as a non-animal alternative to predict absorption. However, these prediction models still have to go far ahead to develop the confidence similar to conventional experimental methods for estimation of drug absorption. Some of the general concerns are selection of appropriate ML methods and validation techniques in addition to selecting relevant descriptors and authentic data sets for the generation of prediction models. The current review explores published models of ML for the prediction of absorption using physicochemical properties as descriptors and their important conclusions. In addition, some critical challenges in acceptance of ML models for absorption are also discussed.

Chemometric modeling of odor threshold property of diverse aroma components of wine.

We have modelled here odor threshold properties (OTP) of various aroma components present in different types of wine using quantitative structure–property relationship (QSPR) studies employing both two-dimensional and three-dimensional descriptors. The aim has been to identify the molecular properties essential for lowering the OTP. We have applied different variable selection strategies to select the most relevant descriptors prior to the development of the final partial least squares (PLS) regression model, which was validated extensively using different validation metrics in terms of acceptability and predictivity of the model for enhancing confidence in QSPR predictions. Using the developed PLS model, we have also predicted the “composite” OTP of different types of wine using the “composite” descriptor values based on individual components according to the PLS model and the results were well corroborated with the observations reported by Wang et al. [Food Chem., 2017, 226, 41–50]. The developed model may guide us to understand the dependence of the odor quality of different types of wines obtained under different manufacturing conditions on their aroma constituents.

Quantitative structure–property relationship approach to predicting xylene separation with diverse exchanged faujasites

Streamlining the xylene separation process on faujasites is a promising way to design innovative adsorbents for this application. For this purpose, we present herein an original quantitative structure-property relationship (QSPR) approach. It deals with the development of a multi-linear predictive model correlating the separation properties with a set of structural descriptors for the adsorbents. The implementation of such an approach makes it necessary to (i) set an appropriate design of experiment (DOE), (ii) prepare an adsorbent database, (iii) test the adsorbent database for xylene separation and (iv) compute a set of relevant descriptors. The selected descriptors essentially characterize the nature of the confinement in the faujasite supercage, i.e., the size of the cations localized in adsorption sites II, as well as the occupancy ratio of both adsorption sites II and III. Two different statistical methods were applied to develop a structure-property relationship model linking experimental selectivity and the set of descriptors. A multiple linear regression model enables the prediction of para/meta-xylene selectivity with a correlation coefficient R2 of 0.78, while a linear discriminant analysis predicts the assignment of the adsorbents to four identified classes with a total prediction percentage of 76%.

A combined Fisher and Laplacian score for feature selection in QSAR based drug design using compounds with known and unknown activities.

Quantitative structure-activity relationship (QSAR) is an effective computational technique for drug design that relates the chemical structures of compounds to their biological activities. Feature selection is an important step in QSAR based drug design to select the most relevant descriptors. One of the most popular feature selection methods for classification problems is Fisher score which aim is to minimize the within-class distance and maximize the between-class distance. In this study, the properties of Fisher criterion were extended for QSAR models to define the new distance metrics based on the continuous activity values of compounds with known activities. Then, a semi-supervised feature selection method was proposed based on the combination of Fisher and Laplacian criteria which exploits both compounds with known and unknown activities to select the relevant descriptors. To demonstrate the efficiency of the proposed semi-supervised feature selection method in selecting the relevant descriptors, we applied the method and other feature selection methods on three QSAR data sets such as serine/threonine-protein kinase PLK3 inhibitors, ROCK inhibitors and phenol compounds. The results demonstrated that the QSAR models built on the selected descriptors by the proposed semi-supervised method have better performance than other models. This indicates the efficiency of the proposed method in selecting the relevant descriptors using the compounds with known and unknown activities. The results of this study showed that the compounds with known and unknown activities can be helpful to improve the performance of the combined Fisher and Laplacian based feature selection methods.

Abiotic Stress Response to As and As+Si, Composite Reprogramming of Fruit Metabolites in Tomato Cultivars.

The toxic element arsenic interacts with the beneficial element silicon at many levels of the plant metabolism. The ability of the tomato plant to take up and translocate As into its fruit has risen concerns that it could facilitate the entry of this element into the human food chain above the admitted level. Here, the fruit of two contrasting tomato cultivars, Aragon and Gladis, were evaluated following exposures of either 48 h or 14 days to As-contaminated irrigation water, with or without supplementary Si. The focus was on selected biochemical stress response indicators to dissect metabolic fruit reprogramming induced by As and Si. A multivariate statistical approach was utilized to establish the relationship between tissue As and Si concentrations and selected biochemical aspects of the stress response mechanisms to identify a set of relevant stress response descriptors. This resulted in the recognition of strong cultivar and temporal effects on metabolic and biochemical stress parameters following the treatments. In this paper the metabolic changes in H2O2 content, lipid peroxidation, lycopene and carotenoids content, ascorbate and GSH redox state, total phenolics, ABTS and DPPH radicals inhibition were in favor of an oxidative stress. The significance of some of these parameters as reliable arsenic exposition biomarkers is discussed in the context of the limited knowledge on the As-induced stress response mechanisms at the level of the ripening fruit which presents a distinctive molecular background dissimilar from roots and shoots.

An materials informatics approach to Ni-based single crystal superalloys lattice misfit prediction

The lattice misfit between γ and γ′ phase in Ni-based single crystal superalloys plays a critical role for microstructural stability and high-temperature creep and fatigue resistance. Making predictions of the lattice misfit rapidly and accurately is therefore of much practical importance, especially for costly and time-consuming material design by trial and error. In this study, we provide a machine learning approach to predict misfit using relevant material descriptors including the chemical composition, dendrite information and measurement temperature and so on. We perform support vector regression, sequential minimal optimization regression and multilayer perceptron algorithms with linear and poly kernels on experimental dataset for appropriate model selecting, and multilayer perceptron model works well for its distinguished prediction performance with high correlation coefficient and low error values. The approach is validated by comparing the predicted lattice misfit with a widely used empirical formula and experimental observation with respect to prediction accuracy.

Reconstruction of membrane current by deconvolution and its application to membrane capacitance measurements in cardiac myocytes.

Correct detection of membrane currents under whole-cell patch–clamp conditions is limited by the transfer function of a recording system. The low-pass output filter of a recording amplifier alters the time course of membrane current and causes errors in relevant descriptors. To solve these problems, we developed a practical procedure for reconstruction of the time course of membrane currents based on deconvolution of recorded currents in frequency domain. The procedure was tested on membrane capacitance estimates from current responses to step voltage pulses. The reconstructed current responses, in contrast to original current records, could be described exactly by an adequate impedance model of a recorded cell. The reconstruction allowed to increase the accuracy and reliability of membrane capacitance measurements in wide range of cell sizes and to suppress the cross-talk errors well below the noise. Moreover, it allowed resolving the instabilities in recording conditions arising from parasitic capacitance and seal resistance variation. Complex tests on hardware models, on simulated data sets, and on living cells confirmed robustness and reliability of the deconvolution procedure. The aptitude of the method was demonstrated in isolated rat cardiac myocytes by recording of spontaneous vesicular events, by discerning the formation of a fusion pore, and by revealing artefacts due to unstable seal resistance.

QSAR Modeling of Aminopeptidase N/CD13 (APN) Inhibitory Activity of some Leucine Ureido Derivatives by GA-MLR and SW-MLR Methods

Background: Aminopeptidase N also known as APN/CD13 is a zinc-dependent type II membrane-bond ectopeptidase that is overexpressed on cancer cells. APN/CD13 is considered as an attractive target for anti-cancer drug design due to key roles in tumor invasion, angiogenesis and metastasis. Therefore, study of quantitative relationships between activity and structures of APN/CD13 inhibitors can provide useful information for the designing and synthesis of novel APN inhibitors. Methods: In this study, the linear method was used to develop QSAR models in order to predict the activities of APN/CD13 inhibitors. A dataset that consisted of 39 leucine ureido derivatives was divided into the training and test subsets. Genetic algorithm and stepwise methods have been employed for selection of relevant descriptors. Results: Multiple linear regression analysis with GA selection was used to model the structureactivity relationships. In this model, R2 was 0.84 for the training set and 0.67 for the test set. We also applied the SW technique as variable selection procedure to the same data set. With the use of SW, the 6 most relevant descriptors were selected to build the model. The value of R2 of the SWMLR model was 0.87 for the training set and 0.77 for the test set. Both models were validated by leave-one-out (LOO) cross-validation. Conclusion: The results of SW-MLR and of GA-MLR were confirmed and is appeared that the SW-MLR model had more power to predict APN inhibitory activity. On the basis of QSAR models, charge polarization, the atomic masses, polarizability, the atomic van der Waals volumes, the molecular symmetry and aromaticity index were found to be important factors controlling the APN inhibitory activity. Keywords: Aminopeptidase, QSAR, multiple linear regressions, stepwise, genetic algorithm, leucine ureido derivatives.

On the issue of the PEMFC operating fault identification: Generic analysis tool based on voltage pointwise singularity strengths

The purpose of this article is to study the portability of a non-intrusive and free of any external/internal disturbance diagnosis tool devoted to the monitoring of the State of Health (SoH) of PEM Fuel Cell (PEMFC) stack. The tool is based on a thorough analysis of the stack voltage signal using a multifractal formalism and wavelet leaders. It offers well-suited signatures indicators on the SoH of the Fuel Cell. Some relevant descriptors extracted from these patterns (singularity features) are used in the frame of Machine Learning approaches to allow the PEMFC fault identification. The proposed diagnosis strategy is evaluated with two different PEMFC stacks. The first one is designed for automotive applications and the second one is dedicated to stationary use (micro combined heat and power - μCHP application). The classification results obtained for the both stacks indicate that the proposed PEMFC diagnosis tool allows identifying simple operating faults as well as more complicated operating situations combining several fault types.

What-and-where to match: Deep spatially multiplicative integration networks for person re-identification

Matching pedestrians across disjoint camera views, known as person re-identification (re-id), is a challenging problem that is of importance to visual recognition and surveillance. Most existing methods exploit local regions with spatial manipulation to perform matching in local correspondences. However, they essentially extract fixed representations from pre-divided regions for each image and then perform matching based on these extracted representations. For models in this pipeline, local finer patterns that are crucial to distinguish positive pairs from negative ones cannot be captured, and thus making them underperformed. In this paper, we propose a novel deep multiplicative integration gating function, which answers the question of what-and-where to match for effective person re-id. To address what to match, our deep network emphasizes common local patterns by learning joint representations in a multiplicative way. The network comprises two Convolutional Neural Networks (CNNs) to extract convolutional activations, and generates relevant descriptors for pedestrian matching. This leads to flexible representations for pair-wise images. To address where to match, we combat the spatial misalignment by performing spatially recurrent pooling via a four-directional recurrent neural network to impose spatial dependency over all positions with respect to the entire image. The proposed network is designed to be end-to-end trainable to characterize local pairwise feature interactions in a spatially aligned manner. To demonstrate the superiority of our method, extensive experiments are conducted over three benchmark data sets: VIPeR, CUHK03 and Market-1501.

Advanced Structure-activity Relationships Applied to Mentha spicata L. Subsp. spicata Essential Oil Compounds as AChE and NMDA Ligands, in Comparison with Donepezil, Galantamine and Memantine – New Approach in Brain Disorders Pharmacology

Alzheimer's disease (AD) therapy is based on several natural and synthetic compounds that act as acetylcholinesterase (AChE) and N-methyl-D-aspartate receptor (NMDA) ligands that have limited efficiency in relieving AD symptoms. Recent studies show that inhibitors isolated from Mentha spicata L. subsp. spicata are promising for AD therapy. We aimed to identify novel and more potent phytopharmaceutical compounds for AD treatment by taking into account the compounds from Mentha spicata L. subsp. spicata essential oil. We generated structure-activity relationship (SAR) models that predict the biological activities of 14 Mentha spicata L. subsp. spicata compounds on AChE and NMDA by comparing their molecular features with those of the three conventional ligands: donepezil, galantamine and memantine. The most relevant descriptors for predicting the biological activities of considered compounds are solvent accessible area and their subdivided, hydrophobicity, energy of frontier molecular orbitals and counts of the aromatic ring and rotatable bounds. 1,8-cineole, the main compound from Mentha spicata L. subsp. spicata essential oil, resulted to be similar with memantine and dissimilar with donepezil in respect to hidrophobicity (logP1,8-cineole=2.95, logPmemantine=2.81, logPdonepezil=4.11), the energy of LUMO (eLUMO1,8-cineole=3.01 eV, eLUMOmemantine=3.35 eV, eLUMOdonepezil=-0.35 eV) and the solvent accessible surface areas over all hydrophobic (SA_H1,8-cineole= 350 Å2, SA_Hmemantine= 358 Å2, SA_Hdonepezil= 655 Å2) or polar atoms (SA_P1,8-cineole= 4 Å2, SA_Pmemantine=10 Å2, SA_Pdonepezil=44.62 Å2). Our results point towards 1,8-cineole as a good candidate for NMDA antagonism, with a weaker AChE inhibitory effect. Our results may be useful in establishing new therapeutic strategies for neurological disorders.

Unraveling the origin of twin related domains and grain boundary evolution during grain boundary engineering

Grain boundary engineering of Fe-based austenitic stainless steels and other materials has been successful in producing a large increase in twin and twin related grain boundaries from a wide range of thermomechanical treatments. However, the exact mechanisms and effective grain boundary network descriptors to create the heavily twinned microstructures are yet to be fully understood. In this study, we provide insight into the grain boundary engineering process by examining sequential progression of the same spatial location of a twin related microstructure through thermomechanical processing. The results show that clusters of twin related grain boundaries called twin related domains form during primary recrystallization. The size of the twin related domains increases as the level of strain falls toward the critical strain for recrystallization. Growth of twin related domains during recrystallization results in the formation of twin boundaries behind the migrating grain boundary front. Formation of higher order twin boundaries occurs when two separate grain boundary fronts of the same twin related domain impinge upon each other. We also present relevant microstructural descriptors with emphasis on twin related domain statistics to recrystallization phenomena in grain boundary engineering materials.

Combining strong sparsity and competitive predictive power with the L-sOPLS approach for biomarker discovery in metabolomics

In the context of metabolomics analyses, partial least squares (PLS) represents the standard tool to perform regression and classification. OPLS, the Orthogonal extension of PLS which has proved to be very useful when interpretation is the main issue, is a more recent way to decompose the PLS solution into predictive components correlated to the target Y and components pertaining to the data X but uncorrelated to Y. This predominance of (O)PLS can raise the question of the awareness of alternative multivariate regression and/or classification tools able to find biomarkers. Actually, the search for biomarkers remains a key issue in metabolomics as it is crucial to very accurately target discriminating features. Most of the time, (O)PLS methods perform well but a drawback often occurs: too many variables can be selected as potential biomarkers even using adapted statistical significance tests. However, for final users (in medical studies for instance), it can be advantageous to deal with only a small number of easily interpretable biomarkers. This drawback is approached in this paper via the use of sparse methods. The sparse-PLS (sPLS), an extension of PLS which promotes an inner variable/feature selection, is an interesting existing solution. But a new intuitive algorithm is proposed in this paper to combine sparsity and the advantages of an orthogonalization step: the “Light-sparse-OPLS” (L-sOPLS). L-sOPLS promotes sparsity on a previously optimized deflated matrix which implies the removal of the Y-orthogonal components. A discussion around the compromise between sparsity and predictive modelling performances is provided and it is shown that L-sOPLS produces convincing results, illustrated principally on the basis of $$^1$$ H-NMR spectral data but also on genomic RT-qPCR data. The L-sOPLS algorithm allows to reach better predictive performances than (O)PLS and sPLS while taking into account only a very small number of relevant descriptors.

Tethered Particle Motion: An Easy Technique for Probing DNA Topology and Interactions with Transcription Factors.

Tethered Particle Motion (TPM) is a versatile in vitro technique for monitoring the conformations a linear macromolecule, such as DNA, can exhibit. The technique involves monitoring the diffusive motion of a particle anchored to a fixed point via the macromolecule of interest, which acts as a tether. In this chapter, we provide an overview of TPM, review the fundamental principles that determine the accuracy with which effective tether lengths can be used to distinguish different tether conformations, present software tools that assist in capturing and analyzing TPM data, and provide a protocol which uses TPM to characterize lac repressor-induced DNA looping. Critical to any TPM assay is the understanding of the timescale over which the diffusive motion of the particle must be observed to accurately distinguish tether conformations. Approximating the tether as a Hookean spring, we show how to estimate the diffusion timescale and discuss how it relates to the confidence with which tether conformations can be distinguished. Applying those estimates to a lac repressor titration assay, we describe how to perform a TPM experiment. We also provide graphically driven software which can be used to speed up data collection and analysis. Lastly, we detail how TPM data from the titration assay can be used to calculate relevant molecular descriptors such as the J factor for DNA looping and lac repressor-operator dissociation constants. While the included protocol is geared toward studying DNA looping, the technique, fundamental principles, and analytical methods are more general and can be adapted to a wide variety of molecular systems.

A systematic evaluation of analogs and automated read-across prediction of estrogenicity: A case study using hindered phenols

Read-across is an important data gap filling technique used within category and analog approaches for regulatory hazard identification and risk assessment. Although much technical guidance is available that describes how to develop category/analog approaches, practical principles to evaluate and substantiate analog validity (suitability) are still lacking. This case study uses hindered phenols as an example chemical class to determine: (1) the capability of three structure fingerprint/descriptor methods (PubChem, ToxPrints and MoSS MCSS) to identify analogs for read-across to predict Estrogen Receptor (ER) binding activity and, (2) the utility of data confidence measures, physicochemical properties, and chemical R-group properties as filters to improve ER binding predictions. The training dataset comprised 462 hindered phenols and 257 non-hindered phenols. For each chemical of interest (target), source analogs were identified from two datasets (hindered and non-hindered phenols) that had been characterized by a fingerprint/descriptor method and by two cut-offs: (1) minimum similarity distance (range: 0.1–0.9) and, (2) N closest analogs (range: 1–10). Analogs were then filtered using: (1) physicochemical properties of the phenol (termed global filtering) and, (2) physicochemical properties of the R-groups neighboring the active hydroxyl group (termed local filtering). A read-across prediction was made for each target chemical on the basis of a majority vote of the N closest analogs. The results demonstrate that: (1) concordance in ER activity increases with structural similarity, regardless of the structure fingerprint/descriptor method, (2) increased data confidence significantly improves read-across predictions, and (3) filtering analogs using global and local properties can help identify more suitable analogs. This case study illustrates that the quality of the underlying experimental data and use of endpoint relevant chemical descriptors to evaluate source analogs are critical to achieving robust read-across predictions.

Clinical descriptors for the recognition of central sensitization pain in patients with knee osteoarthritis

Background: Despite growing awareness of the contribution of central pain mechanisms to knee osteoarthritis pain in a subgroup of patients, routine evaluation of central sensitization is yet to be incorporated into clinical practice.Aim: The objective of this perspective is to design a set of clinical descriptors for the recognition of central sensitization in patients with knee osteoarthritis that can be implemented in clinical practice.Methods: A narrative review of original research papers was conducted by nine clinicians and researchers from seven different countries to reach agreement on clinically relevant descriptors.Results: It is proposed that identification of a dominance of central sensitization pain is based on descriptors derived from the subjective assessment and the physical examination. In the former, clinicians are recommended to inquire about intensity and duration of pain and its association with structural joint changes, pain distribution, behavior of knee pain, presence of neuropathic-like or centrally mediated symptoms and responsiveness to previous treatment. The latter includes assessment of response to clinical test, mechanical hyperalgesia and allodynia, thermal hyperalgesia, hypoesthesia and reduced vibration sense.Conclusions: This article describes a set of clinically relevant descriptors that might indicate the presence of central sensitization in patients with knee osteoarthritis in clinical practice. Although based on research data, the descriptors proposed in this review require experimental testing in future studies.Implications for RehabilitationLaboratory evaluation of central sensitization for people with knee osteoarthritis is yet to be incorporated into clinical practice.A set of clinical indicators for the recognition of central sensitization in patients with knee osteoarthritis is proposed.Although based on research data, the clinical indicators proposed require further experimental testing of psychometric properties.

A Simple, Robust and Efficient Computational Method for n-Octanol/Water Partition Coefficients of Substituted Aromatic Drugs

In this paper, multiple linear regression (MLR) was used to build quantitative structure property relationship (QSPR) of n-octanol-water partition coefficient (logPo/w) of 195 substituted aromatic drugs. The molecular descriptors were calculated for each compound by the VLifeMDS. By applying genetic algorithm/multiple linear regressions (GA/MLR) the most relevant descriptors were selected to build a QSPR model. The robustness of the model was characterized by the statistical validation and applicability domain (AD). The prediction results from MLR are in good agreement with the experimental values. The R2 and Q2LOO for MLR are 0.9433, 0.9341. The AD of the model was analyzed based on the Williams plot. The effects of different selected descriptors are described.

QSAR Study of 17β-HSD3 Inhibitors by Genetic Algorithm-Support Vector Machine as a Target Receptor for the Treatment of Prostate Cancer.

The 17β-HSD3 enzyme plays a key role in treatment of prostate cancer and small inhibitors can be used to efficiently target it. In the present study, the multiple linear regression (MLR), and support vector machine (SVM) methods were used to interpret the chemical structural functionality against the inhibition activity of some 17β-HSD3inhibitors. Chemical structural information were described through various types of molecular descriptors and genetic algorithm (GA) was applied to decrease the complexity of inhibition pathway to a few relevant molecular descriptors. Non-linear method (GA-SVM) showed to be better than the linear (GA-MLR) method in terms of the internal and the external prediction accuracy. The SVM model, with high statistical significance (R2train = 0.938; R2test = 0.870), was found to be useful for estimating the inhibition activity of 17β-HSD3 inhibitors. The models were validated rigorously through leave-one-out cross-validation and several compounds as external test set. Furthermore, the external predictive power of the proposed model was examined by considering modified R2 and concordance correlation coefficient values, Golbraikh and Tropsha acceptable model criteria's, and an extra evaluation set from an external data set. Applicability domain of the linear model was carefully defined using Williams plot. Moreover, Euclidean based applicability domain was applied to define the chemical structural diversity of the evaluation set and training set.