Editing AI for Genomic Medicine (section)

== <span style="color: #FFFFFF;">Analyzing</span> ==
{| class="wikitable"
|+ Genomic Medicine AI Applications
! Application !! Clinical Maturity !! Key AI Model !! Evidence Quality
|-
| Variant pathogenicity (coding) || Clinical use || AlphaMissense, REVEL || Strong
|-
| Pharmacogenomics CDS || Deployed (many hospitals) || Rule + ML hybrid || Strong (RCTs)
|-
| Tumor biomarker detection || Clinical (NGS panels) || DL on sequencing data || Strong
|-
| Multi-cancer early detection || Clinical trials || ML on ctDNA methylation || Moderate
|-
| Polygenic risk score (preventive) || Research → clinical || Penalized regression || Moderate
|-
| Rare disease diagnosis AI || Growing || Phenotype + genotype ML || Moderate
|}

'''Failure modes''': Ancestry bias in variant databases (ClinVar over-represented European ancestry). VUS reclassification errors — incorrectly classifying a pathogenic variant as benign. Tumor heterogeneity — liquid biopsy may miss sub-clonal mutations. Incidental findings — genomic sequencing reveals clinically significant variants unrelated to the reason for testing; consent and return-of-results protocols required. Off-label treatment matching — AI recommending treatments without RCT evidence in that tumor type.
</div>

<div style="background-color: #483D8B; color: #FFFFFF; padding: 20px; border-radius: 8px; margin-bottom: 15px;">