Skip to content

Latest commit

 

History

History
14 lines (9 loc) · 771 Bytes

README.md

File metadata and controls

14 lines (9 loc) · 771 Bytes

Toxicology-Prediction

A model that uses the tox21 dataset [1] to attempt to find a realtionship between molecular structure and realtive toxicity, and a streamlit website to use to interact with it.

This is currently still an ongoing project for the University of Victoria's AI club.

What still needs to be completed:

  • Normalize the dataset using SKlearn's standard scalar
  • Change the way the model trains by altering validation splits and sample wieghts
  • Inclusion of the Drugbank api to convert user-inputed molecular names into SMILE format (Streamlit site)

The dataset has already been modified to include molecular descriptors using mordred, however futher testing with other featurisation methods may be beneficial.

[1] https://tox21.gov/resources/