Back

Learning phenotypic patterns in genetic disease by symptom interaction modeling

Yauy, K.; Duforet-Frebourg, N.; Testard, Q.; Beaumeunier, S.; Audoux, J.; Simard, B.; Larue, D.; Blum, M.; Bernard, V.; Bertrand, D.; Genevieve, D.; PhenoGenius consortium, ; Philippe, N.; Thevenon, J.

2022-07-31 genetic and genomic medicine
10.1101/2022.07.29.22278181 medRxiv
Show abstract

Observing phenotyping practices from an international cohort of 1,686 cases revealed heterogeneity of phenotype reporting among clinicians. Heterogeneity limited their exploitation for diagnosis as only 43% of symptom-gene associations in the cohort were available in public databases. We developed a symptom interaction model that summarized 16,600 terms into 390 groups of interacting symptoms and detected 3,222,053 novel symptom-gene associations. By learning phenotypic patterns in genetic diseases, symptom interaction modeling handled heterogeneity in phenotyping, to the extent of covering 98% of our cohorts symptom-gene associations. Using these symptom interactions improved the diagnostic performance in gene prioritization by 42% (median rank 80 to 41) compared to the best algorithms. Symptom interaction modeling will provide new discoveries in precision medicine by standardizing clinical descriptions. One sentence summaryLearning phenotypic patterns in genetic disease by symptom interaction modeling addresses physicians heterogeneous phenotype reporting.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.