Package: PONG2 1.0.1

Suraju A. Sadeeq

PONG2: KIR Genotype Imputation and Model Training from SNP Array Data

A scalable and accurate tool for Killer-cell Immunoglobulin-like Receptor (KIR) genotype imputation directly from SNP array data using supervised machine learning models trained across five continental ancestry groups. Uses attribute bagging and an ensemble classifier method with haplotype inference for SNPs and KIR types. Models are built from global populations in the 1000 Genomes Project and validated across diverse biobank cohorts. Methods are based on Zheng et al. (2014) <doi:10.1016/j.ajhg.2013.12.015> and Sadeeq et al. (2026) <https://github.com/NormanLabUCD/PONG2>.

Authors:Suraju A. Sadeeq [aut, cre], Laura A. Leaton [aut], Katherine M. Kichula [aut], Paul J. Norman [aut], Xiuwen Zheng [ctb, cph]

PONG2_1.0.1.tar.gz
PONG2_1.0.1.tar.gz(r-4.7-arm64)PONG2_1.0.1.tar.gz(r-4.7-x86_64)PONG2_1.0.1.tar.gz(r-4.6-arm64)PONG2_1.0.1.tar.gz(r-4.6-x86_64)
PONG2_1.0.1.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION
card.svg |card.png
PONG2/json (API)

# Install 'PONG2' in R:
install.packages('PONG2', repos = c('https://cran.r-universe.dev', 'https://cloud.r-project.org'))

Bug tracker:https://github.com/normanlabucd/pong2/issues

Pkgdown/docs site:https://normanlabucd.github.io

Uses libs:
  • c++– GNU Standard C++ Library v3
Datasets:

On CRAN:

Conda:

cpp

3.30 score 10 scripts 53 exports 2 dependencies

Last updated from:ef287ae6d1. Checks:6 OK. Indexed: yes.

TargetResultTimeFilesSyslog
linux-devel-arm64OK173
linux-devel-x86_64OK163
source / vignettesOK273
linux-release-arm64OK158
linux-release-x86_64OK145
wasm-releaseOK167

Exports:.get_model_path.Last.libhlaAllelehlaAlleleDigithlaAlleleSubsethlaAttrBagginghlaBED2GenohlaCheckSNPshlaClosehlaCombineAllelehlaCombineModelObjhlaCompareAllelehlaErrMsghlaGDS2GenohlaGeno2PEDhlaGenoAFreqhlaGenoCombinehlaGenoLDhlaGenoMFreqhlaGenoMRatehlaGenoMRate_SamphlaGenoSubsethlaGenoSubsetFlankhlaGenoSwitchStrandhlaHaplo2GenohlaHaploSubsethlaLociInfohlaMakeSNPGenohlaMakeSNPHaplohlaModelFileshlaModelFromObjhlaModelToObjhlaOutOfBaghlaPredMergehlaPublishhlaReporthlaSampleAllelehlaSNPIDhlaSplitAllelehlaSubModelObjhlaUniqueAllelekirParallelAttrBaggingkirPredictplot.hlaAttrBagClassplot.hlaAttrBagObjpredict.hlaAttrBagClassprint.hlaAttrBagClassprint.hlaAttrBagObjsummary.hlaAlleleClasssummary.hlaAttrBagClasssummary.hlaAttrBagObjsummary.hlaSNPGenoClasssummary.hlaSNPHaploClass

Dependencies:RcppRcppParallel

PONG2 Basics: Installation, Quick Start, and Core Usage
Overview | Features | Requirements | Installation | From GitHub (recommended — latest version) | From release tarball | CLI Setup | Verify installation | Quick Start Examples | 1. Basic imputation | 2. Imputation with missing SNP fill-in | 3. Training a new model | 4. Evaluating a trained model | Core Usage Reference | Help | impute command | Required flags | Optional flags | train command | KIR file format | evaluate command | Pre-phasing the KIR Region | hg19 | hg38 | Improving Imputation Accuracy | Option A: Local pre-imputation (built-in, quick) | Option B: External pre-imputation (recommended for highest accuracy) | Option C: Force imputation (not recommended) | Next Steps
Rendered fromPONG2-basics.Rmdusingknitr::rmarkdown

Last update: 2026-06-24
Started: 2026-06-24

PONG2 Imputation Workflow
Overview | Prerequisites | Step 1: Prepare Input Data | Step 2: Run Basic PONG2 Imputation | Step 3: Check SNP Overlap | Step 4: Pre-imputation (when SNP overlap < 50%) | Pre-phase with Eagle2 | hg19 | hg38 | Option A: Local Pre-imputation with minimac4 (built-in) | Option B: External Pre-imputation (recommended for highest accuracy) | Step B1: Export phased VCF | Step B2: Upload to Michigan Imputation Server | Step B3: Download and convert imputed VCF to PLINK | Step B4: Run PONG2 on imputed data | Option C: Force imputation (not recommended) | Step 5: Interpreting Output | Output CSV format | Large sample datasets | Summary: Which Workflow to Choose? | Next Steps
Rendered fromPONG2-imputation.Rmdusingknitr::rmarkdown

Last update: 2026-06-24
Started: 2026-06-24

PONG2 R API: Direct R Usage with Example Data
Overview | 1. Installation and Setup | 2. Example Data | 3. KIR Genotype Prediction | 4. Model Training | 5. Model Evaluation | 6. CLI Usage | Session Info
Rendered fromPONG2-R-api.Rmdusingknitr::rmarkdown

Last update: 2026-06-24
Started: 2026-06-24

PONG2 Training: Building Custom KIR Prediction Models
Overview | Prerequisites | Step 1: Prepare Input Data | 1a. Reference genotypes (--bfile) | Using the 1000 Genomes Project (1KGP) as reference panel | Using your own reference dataset | 1b. Known KIR allele calls (--kfile) | Format | Rules | Step 2: Run Training | With optional parameters | Key training parameters | Step 3: Training Output | Step 4: Evaluate Model Performance | Option A: Evaluate from the terminal (recommended) | Option B: Evaluate in R | Step 5: Use a Custom Model for Imputation | Troubleshooting | Next Steps
Rendered fromPONG2-training.Rmdusingknitr::rmarkdown

Last update: 2026-06-24
Started: 2026-06-24

Readme and manuals

Help Manual

Help pageTopics
Train KIR prediction models in parallelkirParallelAttrBagging
Predict KIR genotypes from SNP datakirPredict
PONG2 Example Datasetexample_kir example_mobj example_snp PONG2_example