Background: Quantitative Structure-Activity Relationships (QSAR) is a well-established
branch of computational chemistry. The presence of QSAR papers is decreasing for the last few years.
Objective: To highlight recent trends of QSAR in general and trends of machine learning methods in
Method: A bibliometric analysis of articles published in top ten molecular modeling and medicinal
chemistry journals was carried out. The bibliometric statistics was collected for papers published in
2009 and 2015 and compared.
Results: During 5-year span studied, the fraction of QSAR studies underwent a twofold decrease. Top
journals of both categories became less likely to publish Multiple Linear Regression models and
increased the presence of Random forest and Naïve Bayes methods. 3D-QSAR remains the most
popular method of studying structure-activity relationships with a slight decrease of its presence in
molecular modeling journals but a relative increase in medicinal chemistry.
Conclusion: The downward QSAR trend might have several reasons: more stringent criteria for QSAR
studies acceptance by journals, transformation of QSAR studies into routine work due to wider
availability of QSAR methods and the overall maturation of QSAR field, and possible disappointment in
QSAR. We expect that the progress in machine learning methods being adopted by
chem(o)informaticians finally will help QSAR to find its place in drug design and to move to the
Plateau of Productivity.