New statistical model effectively predicts the toxicity of pharmaceutical molecules
A statistical component model can be used to identify associations between chemicals and toxicological effects.
The joint Aalto University, Karolinska Institute and Institute for Molecular Medicine Finland (FIMM) study included over 1,300 known pharmaceutical molecules, on which there is a wealth of measurement data available.
‘The study uses systematic data-driven analysis to combine toxicity measurements taken on cell lines with gene expression responses describing gene activation. Toxicity includes growth inhibitory and cell killing effects. The method developed in the study makes it possible to more accurately predict the toxicity of new molecules because it makes use of advanced statistical methods and much bigger datasets than before,’ explains Juuso Parkkinen, who completed his doctoral dissertation at Aalto University.
At present, toxicity is primarily measured by means of animal testing. Thanks to this new method, animal testing can be partly replaced in the future by a combination of cell line testing and statistical modelling. This would also result in considerable cost savings for pharmaceutical development.
‘The new prediction method can be applied to new pharmaceutical molecules and other chemicals currently in product development to eliminate possible toxic molecules,’ adds Parkkinen.
Advances in statistical machine learning and artificial intelligence methods have risen to play a crucial role in many application areas in addition to medical research.
‘Juuso Parkkinen is an excellent example of the usefulness of Aalto University's artificial intelligence research and doctoral studies: He wrote his dissertation on medicinal applications in my research group and then transferred to Reaktor to apply data science to a wide range of business needs,’ praises Parkkinen's dissertation adviser, Professor Samuel Kaski.
Pharmaceutical toxicity was studied by Juuso Parkkinen and Samuel Kaski of Aalto University, Pekka Kohonen, Egon Willighagen, Rebecca Ceder and Roland Grafström of Karoliska Institute and Krister Wennerberg of the Institute for Molecular Medicine Finland (FIMM).
Further information:
Juuso Parkkinen
AI Designer and Senior Data Scientist
Reaktor
juuso.parkkinen@reaktor.fi
tel. +358 50 356 3916
Samuel Kaski
Professor
Aalto University
samuel.kaski@aalto.fi
tel. +358 50 305 8694
Journal article in Nature Communications: A transcriptomics data-driven gene space accurately predicts liver cytopathology and drug-induced liver injury