29af14b47208040441ab4ff7acf51aac71cad4e2
lrnassar
  Thu Sep 17 21:58:18 2026 -0700
Adding MaveMD track, clinically calibrated multiplexed variant effect measurements on hg38. refs #37800

MaveMD is a curated collection inside MaveDB holding the score sets with clinical
relevance, restricted to genes with a moderate or stronger gene-disease association.
Data come from the MaveDB API rather than the Zenodo snapshots, at the request of the
MaveDB team, who also set the request pacing the fetch script follows.

Two views under a new phenDis container. mavemdVar draws one item per measured variant
per score set, carrying the assay score, the functional class, the ACMG/AMP functional
evidence code and strength where the score set is calibrated, the assay metadata, and
matching ClinVar and gnomAD annotations. mavemdMap draws each score set as a variant
effect map, one column per amino acid position and one row per substitution, reusing
Jonathan's heatmap display from the MaveDB track.

MaveDB resolves about a third of the collection to the genome and the rest only to a
protein sequence, so variants are placed from the genomic HGVS term where one exists,
otherwise by projecting the protein term onto its codon through ncbiRefSeqLink and
ncbiRefSeqCurated for RefSeq accessions or the GENCODE tables for Ensembl ones,
otherwise by running the submitted transcript term through hgvsToVcf. The projection is
cross-checked against the variants carrying both coordinate systems and the build aborts
if they disagree beyond a threshold. Codons split across an exon junction are written as
BED12 with the two real blocks. Haplotypes cannot be given a single position and are
excluded, with every dropped measurement counted by reason in the makeDoc.

Also adds reciprocal relatedTracks entries between mavedb and mavemd. Gated alpha
pending QA.

diff --git src/hg/makeDb/trackDb/relatedTracks.ra src/hg/makeDb/trackDb/relatedTracks.ra
index 024fe07cadc..451f1cb37c5 100644
--- src/hg/makeDb/trackDb/relatedTracks.ra
+++ src/hg/makeDb/trackDb/relatedTracks.ra
@@ -599,30 +599,34 @@
 hg38 spliceAI spliceVarDb Splicing variants with experimental validation, useful for checking these predictions
 hg38 spliceVarDb spliceAI Predicted splice-altering effects, scored genome-wide
 hg38 spliceAI spliceAIWt The same model run on the reference sequence, showing where splicing is expected without any variant
 hg38 spliceAIWt spliceAI The same model scored for variants, rather than for the reference sequence
 hg38 abSplice spliceVarDb Splicing variants with experimental validation, useful for checking these predictions
 hg38 spliceVarDb abSplice Predicted aberrant splicing, scored per variant and tissue
 hg38 spliceImpactSuper predictionScoresSuper Pathogenicity scores for coding and non-coding variants generally, not only splicing
 hg38 predictionScoresSuper spliceImpactSuper Prediction scores and databases for variants that disrupt splicing
 hg38 nmd spliceImpactSuper Predicted and validated splice-altering variants, a common source of premature termination codons
 hg38 spliceImpactSuper nmd Regions where premature termination codons are predicted to escape nonsense-mediated decay
 
 hg19 ~spliceAI abSplice Another deep-learning predictor of splice-altering variants
 hg19 spliceImpactSuper predictionScoresSuper Pathogenicity scores for coding and non-coding variants generally, not only splicing
 hg19 predictionScoresSuper spliceImpactSuper Prediction scores and databases for variants that disrupt splicing
 
+# MaveDB / MaveMD cross-links:
+hg38 mavedb mavemd The clinically curated subset, with ACMG functional evidence calibrations
+hg38 mavemd mavedb The full set of variant effect maps in MaveDB, without clinical calibration
+
 # Constraint score cross-links:
 hg38 constraintSuper predictionScoresSuper Per-variant deleteriousness and pathogenicity scores, rather than regional constraint
 hg38 predictionScoresSuper constraintSuper Regional and gene-level constraint measured from population variation
 hg38 ~jarvis ukbDepletion Another score for how depleted of variation a non-coding region is
 hg38 ~hmc gnomadPLI Another constraint metric derived from the absence of variation in population data
 hg38 hmc ucscGenePfam The Pfam protein domains that homologous missense constraint is calculated over
 hg38 ucscGenePfam hmc Missense constraint measured across homologous positions within these domains
 hg38 promoterAi jarvis A score prioritizing non-coding regions more broadly, not only promoters
 hg38 jarvis promoterAi A deep-learning predictor for variants in promoter regions specifically
 
 hg19 constraintSuper predictionScoresSuper Per-variant deleteriousness and pathogenicity scores, rather than regional constraint
 hg19 predictionScoresSuper constraintSuper Regional and gene-level constraint measured from population variation
 hg19 ~jarvis ukbDepletion Another score for how depleted of variation a non-coding region is
 hg19 ~hmc gnomadPLI Another constraint metric derived from the absence of variation in population data
 hg19 hmc ucscGenePfam The Pfam domains that homologous missense constraint is calculated over