Coherent Disease Subtyping Dashboard

We built two tools based on existing literature. The first, for clinical diagnostics, estimates the probability that any given sample in a colorectal cancer cohort is subtyped correctly. The second, for translational researchers and clinicians, suggests potential therapeutic avenues given a colorectal cancer type and variant analysis.

Given available validity data, this interactive tool estimates the likelihood of correct subtyping in colorectal cancer, given the Guinney, et al classification systme (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4636487/), and the data obtained by a number of research or diagnostic datasets. References for validity measurement of each datatype and combinations thereof are available in the supplementary information file contained in this repo . It is our hope that this approach -- and code -- can be used to build such interactive tools for many diseases, as well as advise diagnostic groups on how to set up validity assessments.

Given available drug efficacy data, this interactive tool recommends extant approved drug that may be appropriate for subjects given their CRC subtype classification, and variant information where available. References for trial data and gene/pathway associations are available in the supplementary information file contained in this repo . It is our hope that this approach -- and code -- can be used to build such interactive tools for many diseases.

Ref: Curr Gastroenterol Rep. 2019 Jan 30; 21(2): 5.

Nat Med. 2015 Nov; 21(11): 1350–1356.

Cancers (Basel). 2021 Oct; 13(19): 4923.

https://nick-giangreco.shinyapps.io/cds-dashboard/

We are able to predict appropriate clinical diagnostic combinations that maximize the accuracy of potential treatment options. Concurrently, we would like to be able to generate an information density graph (radar plot) for putative disease subtypes, highlighting pathways and variants that may be especially relevant to subjects within them. It is our hope that this work will continue over a number of hackathons as well as more pedestrian software development engagements after oncologist review.

Emerson Huitt Nick Giangreco Ames Ma Anthony Costa

Vivian Neilley Nuria Queralt Rosinach Jerven Bolleman Aina Jene

Ben Busby [email protected]