CADD v1.7: using protein language models, regulatory CNNs and other nucleotide-level scores to improve genome-wide variant predictions
Problem
Genome-wide prioritization of genetic variants across molecular functions is needed to aid assessment of clinical findings and genetic studies.
Why it matters
Variant prioritization directly supports clinical interpretation of genetic findings.
What the AI did
CADD v1.7 integrates protein language model scores (Meta ESM-1v), regulatory CNN variant effect predictions, and Zoonomia conservation scores into its ML-based variant scoring, evaluated on ClinVar, ExAC/gnomAD, 1000 Genomes and 31 ProteinGym DMS datasets.
Results
not stated in source
Impact
CADD is stated to be employed to prioritize variants in diverse genetic studies and analyses.
Record
| Venue | Nucleic Acids Research |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Max Schubach, Thorben Maaß, Lusiné Nazaretyan, Sebastian Röner, Martin Kircher |
| Citations | 440 (OpenAlex, at import time) |
| References | 101 |
| Model or method | Meta ESM-1v protein language model; sequence-based CNNs |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1093/nar/gkad989 |
| Open access | gold |
| Keywords | variant prioritization, CADD, protein language models, CNNs, genomics, machine learning |
Importance rating
significant — thousands to millions affected
Abstract states variant scoring aids clinical assessment and is employed in diverse genetic studies; foundational genomics tool, but no population scale given.