predict¶
Generate genomic predictions from fitted SNP effects.
Use this command after fit to score target samples.
Basic Syntax¶
gelex predict -b target_data -e model.snp.eff -o target.pred.tsv
gelex predict --bfile <genotype_prefix> --snp-eff <snp_effect_file> --out <prediction_file> [OPTIONS]
Required inputs are genotype prefix (--bfile), SNP effects
(--snp-eff), and output path (--out).
Options¶
Quick Start Options
-b, --bfilerequiredPLINK binary prefix (
.bed/.bim/.fam) for target samples.-e, --snp-effrequiredSNP effects file from
gelex fit(usually<out>.snp.eff).-o, --outrequiredOutput file path for prediction results.
Input and Covariate Options
--covar-effOptional covariate effect file (usually
<out>.paramfromfit).--qcovarQuantitative covariate TSV in format
FID IID covar1 ....--dcovarCategorical covariate TSV in format
FID IID factor1 ....
Processing Options
--iid-onlyfalseMatch samples by IID only and ignore FID.
-c, --chunk-size10000Number of SNPs per processing chunk. Lower values reduce peak memory.
Output Files¶
predict writes one prediction table at the exact path passed to --out.
File |
Contents |
Reference |
|---|---|---|
|
Per-sample predictions (FID/IID and predicted values) |
Warnings and Notes¶
Warning
Use --iid-only only when IID uniquely identifies individuals in every
file. Otherwise, samples may be mismatched.
Note
If you use --covar-eff, keep covariate files consistent with the fit
stage (same variables, coding, and compatible sample IDs).
Examples¶
gelex predict \
-b target_data \
-e trained_model.snp.eff \
-o predictions.pred.tsv
gelex predict \
-b target_data \
-e trained_model.snp.eff \
--covar-eff trained_model.param \
--qcovar target_age.tsv \
--dcovar target_sex.tsv \
-o predictions_with_covar.pred.tsv
gelex predict \
-b target_data \
-e trained_model.snp.eff \
--iid-only \
-o predictions_iid_only.pred.tsv
gelex predict \
-b target_data \
-e trained_model.snp.eff \
--chunk-size 2000 \
-o predictions_low_mem.pred.tsv
See Also¶
fit for training SNP and covariate effects.
Prediction Results (.pred.tsv) for prediction output columns.