Dataset Information

Systematic artifacts in support vector regression-based compound potency prediction revealed by statistical and activity landscape analysis.

ABSTRACT: Support vector machines are a popular machine learning method for many classification tasks in biology and chemistry. In addition, the support vector regression (SVR) variant is widely used for numerical property predictions. In chemoinformatics and pharmaceutical research, SVR has become the probably most popular approach for modeling of non-linear structure-activity relationships (SARs) and predicting compound potency values. Herein, we have systematically generated and analyzed SVR prediction models for a variety of compound data sets with different SAR characteristics. Although these SVR models were accurate on the basis of global prediction statistics and not prone to overfitting, they were found to consistently mispredict highly potent compounds. Hence, in regions of local SAR discontinuity, SVR prediction models displayed clear limitations. Compared to observed activity landscapes of compound data sets, landscapes generated on the basis of SVR potency predictions were partly flattened and activity cliff information was lost. Taken together, these findings have implications for practical SVR applications. In particular, prospective SVR-based potency predictions should be considered with caution because artificially low predictions are very likely for highly potent candidate compounds, the most important prediction targets.

SUBMITTER: Balfer J

PROVIDER: S-EPMC4350943 | biostudies-literature | 2015

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Systematic artifacts in support vector regression-based compound potency prediction revealed by statistical and activity landscape analysis.

Balfer Jenny J Bajorath Jürgen J

PloS one 20150305 3

Support vector machines are a popular machine learning method for many classification tasks in biology and chemistry. In addition, the support vector regression (SVR) variant is widely used for numerical property predictions. In chemoinformatics and pharmaceutical research, SVR has become the probably most popular approach for modeling of non-linear structure-activity relationships (SARs) and predicting compound potency values. Herein, we have systematically generated and analyzed SVR prediction ...[more]

PMID: 25742011

Similar Datasets

Project description:BackgroundPlasminogen (Pg), the precursor of the proteolytic and fibrinolytic enzyme of blood, is converted to the active enzyme plasmin (Pm) by different plasminogen activators (tissue plasminogen activators and urokinase), including the bacterial activators streptokinase and staphylokinase, which activate Pg to Pm and thus are used clinically for thrombolysis. The identification of Pg-activators is therefore an important step in understanding their functional mechanism and derives new therapies.MethodsIn this study, different computational methods for predicting plasminogen activator peptide sequences with high accuracy were investigated, including support vector machines (SVM) based on amino acid (AC), dipeptide composition (DC), PSSM profile and Hybrid methods used to predict different Pg-activators from both prokaryotic and eukaryotic origins.ResultsOverall maximum accuracy, evaluated using the five-fold cross validation technique, was 88.37%, 84.32%, 87.61%, 85.63% in 0.87, 0.83,0.86 and 0.85 MCC with amino (AC) or dipeptide composition (DC), PSSM profile and Hybrid methods respectively. Through this study, we have found that the different subfamilies of Pg-activators are quite closely correlated in terms of amino, dipeptide, PSSM and Hybrid compositions. Therefore, our prediction results show that plasminogen activators are predictable with a high accuracy from their primary sequence. Prediction performance was also cross-checked by confusion matrix and ROC (Receiver operating characteristics) analysis. A web server to facilitate the prediction of Pg-activators from primary sequence data was implemented.ConclusionThe results show that dipeptide, PSSM profile, and Hybrid based methods perform better than single amino acid composition (AC). Furthermore, we also have developed a web server, which predicts the Pg-activators and their classification (available online at http://mamsap.it.deakin.edu.au/plas_pred/home.html). Our experimental results show that our approaches are faster and achieve generally a good prediction performance.

Dataset Information

Systematic artifacts in support vector regression-based compound potency prediction revealed by statistical and activity landscape analysis.

Publications

Systematic artifacts in support vector regression-based compound potency prediction revealed by statistical and activity landscape analysis.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets