Predicting The Effects of Chemical-Protein Interactions On Proteins Using Tensor Factorisation

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Understanding chemical-protein interactions is crucial for drug design and elucidating drug side-effects.
• Current computational methods predict interactions but not their downstream effects on proteins, such as expression or abundance changes.
• There is a need for advanced computational approaches to model the multifaceted implications of chemical-protein interactions.

Purpose of the Study:

• To develop a computational method for modeling the effects of chemical-protein interactions on protein activity.
• To utilize 3D tensors for representing chemicals, proteins, and their interaction effects.
• To predict the diverse impacts of chemical substances on human proteins.

Main Methods:

• Proposed a novel approach using 3D tensors to model chemicals, target proteins, and associated interaction effects.
• Employed multi-part embedding tensor factorization for predicting chemical effects on human proteins.
• Assessed predictive accuracy using a newly constructed benchmark dataset.

Main Results:

• The proposed 3D tensor factorization method demonstrated strong predictive accuracy.
• Computational experiments confirmed the superior performance of the new approach compared to existing tensor factorization methods.
• Successfully predicted various effects of chemical-protein interactions on protein activity.

Conclusions:

• The 3D tensor modeling and multi-part embedding tensor factorization approach effectively predicts chemical effects on human proteins.
• This method offers valuable insights for drug repurposing and identifying potential drug side-effects.
• The developed approach advances computational strategies in chemical-protein interaction effect prediction.