phenix imputes missing entries in a phenotype matrix using a low-rank multi-trait mixed model and a genetic similarity matrix (R package, Dahl et al 2016 NG). For large datasets, it is better to use methods like softImpute (Mazumder et al 2010 JMLR, Dahl et al 2023 NG) or AutoComplete (An et al 2023 NG)

rgwas finds and validates subtypes in multitrait data by (1) clustering phenotypes using a mixture of regressions (MFMR) and (2) testing for subtype-specific genetic and nongenetic effects. MFMR is essentially a Gaussian mixture model that accounts for population structure bias. (R package, Dahl et al 2019 PG)

gxemm fits polygenic gene-context interaction  (R package). Applies to any low-dimensional "context," e.g. stress (Dahl et al 2020 AJHG) or treatment status (Sadowski et al 2024 biorxiv). GENIE dramatically scales up GxEMM's IID model (Pazoki et al 2023 biorxiv); for binary traits or richer GxE models, you can apply GxEMM to substs and meta-analyze.

singher estimates heritability explained by singleton variants. singher cannot fit standard genetic relatedness matrices (R package, Hernandez et al 2019 NG)

EFA fits the Epistasis Factor Analysis model, a coordinated form of epistasis. EFA scales to large biobanks and jointly fits dozens-hundreds of SNPs (python package, R script, Tang et al 2023 AJHG)

CTMM learns cell type-specific and -shared variation across individuals from single cell RNA-seq data fit on dozens-hundreds of individuals and several cell types (python package, Chen and Dahl 2024 NC)