https://www.nature.com/articles/s41586-021-03922-4
https://twitter.com/SEHanlon/status/1513548800949989377
At #AACR22, @VanAllenLab gives a nice overview of his paper using an interpretable NN model to get insights into determining cancer severity (https://nature.com/articles/s41586-021-03922-4)
Biologically informed deep neural network for prostate cancer discovery
Liked the way he hard-coded specific genes & pathways into the model & looked in detail where the model misclassified specific patients #AACR22 #AACR2022
Also, thought the hard-coding of genes into the model was similar to that in another interpretable AI approach (for brain disease, https://science.org/doi/10.1126/science.aat8464) Could have used the “rank projection trees” from this to highlight important genes #AACR22 #AACR2022
linkstream2 microblog 