MAGPIE: accurate pathogenic prediction for multiple variant types using machine learning approach
Problem
Identifying pathogenic variants among the vast majority of human nucleotide variation remains a challenge.
Why it matters
Distinguishing disease-causing from benign variation is fundamental to interpreting human genome sequencing data.
What the AI did
MAGPIE is a machine learning tool trained on ClinVar that predicts pathogenicity across multiple variant types, performing best on rare variants and highly imbalanced datasets.
Results
Superior performance on an independent test set and multiple orthogonal validation datasets; best on rare variants and imbalanced data; no numeric metrics stated.
Impact
Peer-reviewed publication with 31 citations; tool publicly released on GitHub.
Record
| Venue | Genome Medicine |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Yicheng Liu, Tianyun Zhang, Ningyuan You, Sai Wu, Ning Shen |
| Citations | 31 (OpenAlex, at import time) |
| References | 39 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1186/s13073-023-01274-4 |
| Open access | gold |
| Keywords | variant pathogenicity, machine learning, ClinVar, MAGPIE, genome variation |
Importance rating
significant — thousands to millions affected
Distinguishing pathogenic from benign variants is fundamental to interpreting human genome sequencing data, a foundational genomics task; abstract gives no scale numbers.