Lessons from the CAGI-4 Hopkins clinical panel challenge

Area of Science:

• Genomics
• Clinical Diagnostics
• Bioinformatics

Background:

• Assessing the accuracy of clinical phenotype prediction from DNA sequence is crucial for genetic diagnostics.
• The Johns Hopkins DNA Diagnostic Laboratory provided patient exonic sequences for this evaluation.
• Existing diagnostic methods may miss clinically relevant variants, especially with targeted gene panels.

Purpose of the Study:

• To evaluate state-of-the-art methods for clinical phenotype prediction using DNA sequence data.
• To assess the performance of different prediction groups in identifying disease classes and causal variants.
• To understand the implications for assessing variants of unknown significance and the false-positive rate of DNA-guided analysis.

Main Methods:

• Participants analyzed exonic sequences of 83 genes from 106 patients.
• Five groups predicted disease class probabilities and causal variants.
• Performance was evaluated against diagnoses from the Johns Hopkins DNA Diagnostic Laboratory.

Main Results:

• At least one predictor correctly identified the disease class in 84% of patients with reported variants (36/43).
• Predictors achieved 62% accuracy (39/63) even when the lab did not find a variant.
• Each group uniquely diagnosed at least one patient, highlighting complementary diagnostic capabilities.

Conclusions:

• Computational methods show promise for clinical phenotype prediction from DNA sequence.
• Targeted gene panels may lead to missed diagnoses; broader analysis could be beneficial.
• The study quantifies the false-positive rate in DNA-guided analysis without prior phenotypic information.