• assigned: felixl, alexm, ai-lab
• type: project
• priority: high
• status: ongoing
• keywords: genotyping, smoothing, QTL hotspots, biological signals, noise, markers (SNPs), reverse genotyping

Improve genotype representation for mapping workflows (for example, GEMMA and QTL mapping) by combining classical genotype smoothing with a hotspot-aware "reverse genotyping" strategy.

Genotype datasets are large and often contain redundant marker information. Denser marker sets can increase compute cost and may amplify noise, which can reduce mapping efficiency and interpretability.

Classical smoothing reduces markers by representing recombination intervals with boundary markers (for example, proximal and distal markers). This improves speed, but it may discard non-selected markers that still capture meaningful biological signal.

Reverse genotyping is designed to reduce that risk by using observed QTL hotspot patterns across many traits to guide marker retention.

• [x] Access and prepare BXD 13k traits as the case study dataset.
• [+] Use a subset of 300 traits for testing, then scale to the full set.
• [ ] Build a model that stacks QTL peaks into a single reference profile (`Rev.Geno.Ref`) for BXD traits.
• [+] Examine the QTL trait-file structure.
• [ ] Define hotspot detection criteria (for example, statistical thresholds and overlap rules).
• [x] Generate haplotype data from BXD genotype data.

• [ ] Phase 01: Run mapping for selected traits using only haplotype data.
• [ ] Phase 02: Run mapping for same selected traits using haplotype data enriched with `Rev.Geno.Ref`.
• [ ] Compare Phase 01 vs Phase 02 plots and assess which approach better captures biologically meaningful hotspots. (for justification)

• Step 1: Run trait mapping and collect QTL hits across all available traits (BXD target: ~13k traits). {already in place}
• Step 2: Identify genomic regions where significant peaks overlap across traits (hotspots).
• Step 3: Build a stacked hotspot profile that summarizes shared high-signal regions.
• Step 4: Use this profile to refine marker selection in smoothed haplotype data by keeping markers aligned with hotspot regions and deprioritizing markers repeatedly unsupported by hotspot evidence.
• Step 5: Re-run mapping and evaluate signal quality, speed, and interpretability.

For reference, this single-trait plot illustrates the standard threshold-based interpretation:

• https://genenetwork.org/run_mapping/a539822f0c329d628942e12db93ebfb4?mapping_run_time=5.375121116638184

Example interpretation:

• If the significance threshold is 3.0, the major peak in this example appears on chromosome 7.

The goal is to combine:

• improved mapping accuracy;
• faster runtime from reduced marker complexity;
• clearer hotspot-driven interpretation of trait-marker associations.

How can pangenome resources and AI/ML models improve hotspot detection, marker prioritization, and robustness of the reverse genotyping pipeline?