Anshul Kundaje#1779
The drug response prediction model in Rhaister expresses the response of a drug as a linear combination of responses to panel drugs. By focusing on responses, and using a large amount of training data, we can learn these patterns and make predictions with optimal performance.
Users praise the simplicity of the Rhaister Model for predicting drug responses as linear combinations of panel data.
See also Nima's thread
Read our full manuscript on https://tahoebio-assets.com/rhaister-manuscript.pdf and have a look at our new Emerald Bay dataset at https://huggingface.co/collections/tahoebio/rhaister
@vallens I love the simplicity of the idea. Have you tried reconstructing GRN from those linear models? At least using those drugs with well defined target. For the others the model can likely be used to figure out their target.
Looking forward to read this work in details.
Anshul Kundaje#1779The drug response prediction model in Rhaister expresses the response of a drug as a linear combination of responses to panel drugs. By focusing on responses, and using a large amount of training data, we can learn these patterns and make predictions with optimal performance.