Loading…
This event has ended. Visit the official site or create your own event on Sched.
Welcome to the Rocky 2021 Conference. Please click on the links below to access the Rocky website and the list of posters:

CONFERENCE RESOURCES
Rocky Website
Poster Presentation List without abstracts
Poster Presentation List with abstracts
Back To Schedule
Saturday, December 4 • 10:40am - 10:50am
OP 40 - Inferring Pediatric Sickle Disease Genotypes from Molecular Mechanistic Knowledge

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

OP-40
Inferring Pediatric Sickle Disease Genotypes from Molecular Mechanistic Knowledge

Presenting Author: Tiffany Callahan, University of Colorado Anschutz Medical Campus

Co-Author(s):
Jordan M. Wyrwa, University of Colorado Anschutz Medical Campus
William A. Baumgartner Jr, University of Colorado Anschutz Medical Campus
Lawrence E Hunter, University of Colorado Anschutz Medical Campus
Michael G Kahn, University of Colorado Anschutz Medical Campus

Abstract: Morbidity and mortality from sickle cell disease (SCD) varies widely. Effectively treating SCD requires genotype information. Electronic health records are a valuable source of both individual- and population-level data, but most do not contain genomic data. The objective of this work was to examine whether Med2Mech, a joint learning framework for inferring molecular characterizations of patients from clinical data and publicly available biomedical data, could be used to detect SCD genotypes. Clinical data were obtained for 2,646 pediatric rare disease (816 SCD) and 10,000 control patients from the Children's Hospital of Colorado (CHCO). Genotype data was obtained for 198 (51 HbSC, 147 HbSS) pediatric SCD patients from the Gene Expression Omnibus (GEO). Patient representations built using Med2Mech and Kruskal-Wallis nonparametric ANOVAs were used to determine if the mean rank cosine similarity between the CHCO patient groups to the SCD GEO patients differed. Results revealed that CHCO SCD patients were significantly more similar to GEO patients with their respective genotypes than to other rare disease and control patients (HbSS [n=454]: 2(3)=80,760.30, p<0.001; HbSC [n=347]: 2(3)=27,820.50, p<0.001). Further, using the inferred genotypes assigned by Med2Mech revealed that 14.4% of CHCO SCD patients had at least one potentially erroneous diagnosis and 35.3% had no occurrence of any relevant primary diagnosis. These preliminary findings support using Med2Mech to infer important patient-level data, like genotypes, from publicly available resources, which would otherwise be unavailable.


Presenters
avatar for Tiffany Callahan

Tiffany Callahan

PhD Student, University of Colorado
Computational Biologist, data scientist, and knowledge engineer, interested in pursuing opportunities at the intersection of computer science, natural language processing, and machine learning. My PhD thesis leverages graph representation learning and probabilistic reasoning of biological... Read More →


Saturday December 4, 2021 10:40am - 10:50am MST
Ballroom Salon 1