K-nie/enviGS

Enviromic genomic selection for West African cereals (R package) # enviGS Enviromic genomic selection for drought tolerance in cereals (maize, rice, sorghum), calibrated for West African breeding programs. enviGS takes a breeder from raw marker files to ranked, environment-aware selection decisions. It wraps the trusted R mixed-model stack (sommer, BGLR, rrBLUP) rather than reimplementing it, and reports genotype-by-environment structure through **envirotype-annotated which-won-where biplots** — the classical GGE idiom breeders already read, with the anonymous environment axis replaced by interpretable stress envirotypes. ## Install ```r # install.packages("devtools") devtools::install_github("K-nie/enviGS") ``` Core prediction needs `rrBLUP`; reaction-norm and CV0 use `sommer`; VCF input uses `vcfR`. These are `Suggests` — install the ones you need. ## The pipeline | Stage | Function | |-------|----------| | Ingest genotypes (VCF / HapMap / PLINK / DArTseq / dosage) | `read_geno()` | | Ingest phenotypes, environments, manifest | `read_pheno()`, `read_env()`, `read_manifest()` | | Assemble & reconcile IDs | `build_project()` | | Quality control | `qc_geno()` | | Impute | `impute_geno()` | | Relationship matrices (additive / dominance / enviromic) | `kinship()`, `dominance_kernel()`, `enviromic_kernel()` | | Fit prediction model | `fit_gs()` | | Cross-validate (CV1 / CV2 / CV0) | `cross_validate()` | | Predict breeding values | `predict()` | | Multi-trait selection index | `selection_index()` | | Envirotype which-won-where | `gge_envirotype()`, `plot_biplot()` | | Optimise the testing network | `optimize_network()` | | Reproducible report | `build_report()` | ## Quick start ```r library(enviGS) proj QC -> kinship -> genomic prediction -> envirotype biplot -> multi-trait index -> reproducible report. - Next: in-silico hybrid construction from parents + cross table; reaction-norm G x E via sommer; retrospective validation on DTMA/STMA maize. - Later: a locally-deployable Shiny front end over this same engine (no genotype …