Leveraging deep learning modeling and phenotypic profiling insights for compound bioactivity prediction
The time-consuming and expensive process of experimental compound screening hampers drug discovery. This project aims to develop deep learning models that leverage existing gene expression profiling data to accurately predict the bioactivity of compounds (with rare physical analyses). The LINCS L1000 platform provides cost-effective and scalable cellular disruption data for thousands of compounds. Suitable deep neural network architectures will be evaluated for bioactivity prediction. High-performing models will be rigorously evaluated to demonstrate the viability of this in silico screening approach powered by deep learning and transfer learning.