Cross-Modality Fusion Networks for Bio Activity Regression using Amyotrophic Lateral Sclerosis and Alzheimer’s Data

The current research aims for a precise drug activity regression model through the Cross-Modality Fusion Networks. The model uses different kinds of data to improve its performance. It predicts two important drug activity scores, Half Maximal Inhibitory Concentration and Half Maximal Effective Concentration, which help in developing better treatments for brain diseases like ALS and Alzheimer’s. The different data modalities included are Drug SMILES, protein sequences, Gene expression, and Copy Number variation data. Each type of data is passed to its corresponding algorithms and bilinear transformation. The trials use a well-selected dataset with 665 samples for each branch of the proposed model. After tuning the hyperparameters, the results show that the model performs very well, achieving an R2 score of 0.94, a mean absolute error of 0.17, and a root mean square error of 0.18.