Collaborative Filtering To Repurpose a Drug for COVID

I’m attempting to use collaborative filtering to predict the efficacy of various existing drugs on COVID-19.

Here’s my dataset on github.

And here’s my notebook on github.

My questions are:

  1. Am I going in the right direction in attempting to build the model?
  2. How do I test which drugs will have any kind of an appreciable impact on COVID-19? In other words, given a bunch of rows like so:

COVID-19, chloroquine
COVID-19, ritonavir
COVID-19, lopinavir

I want to show a listing of other drugs from this drug database which are predicted to have some efficacy (greater than some threshold that I define) on COVID-19. How do I show this listing?

I’d appreciate any help, thoughts or ideas.


Remdesivir is what I am studying more deeply coming out of Gilead.

Okay. Thanks to the excellent documentation, I was able to figure out the answers to most of my questions.
Here is my notebook if somebody wants to build on top of it -

I’m hoping that somebody who is well-versed with the fastai api can take a look and provide feedback.