Interpretable Machine Learning on Soybean Multi-Omics Data Reveals Drought-Driven Shifts of Plant–Microbe Interactions

This repository contains all scripts and data used for the analysis in “Interpretable Machine Learning on Soybean Multi-Omics Data Reveals Drought-Driven Shifts of Plant–Microbe Interactions.”

📄 Preprint available at: https://doi.org/10.1101/2025.08.13.670005

---

---

📂 script_ShapPlantMicro

All scripts

File	Description
c0_RF-markers.Rmd	Random Forest marker evaluation
c1_RF_prediction.Rmd	Random Forest prediction
c2_BLUP_prediction.Rmd	BLUP (Best Linear Unbiased Prediction)
c3_GWAS.Rmd	GWAS (Genome-Wide Association Study) analysis
c4_SNP_Mapping.Rmd	SNP selection by RF and GWAS
c5_SHAP_single.Rmd	SHAP analysis for single feature selection
c6_SHAP_interaction.Rmd	SHAP for interaction feature selection
c7_SHAP_difference.Rmd	SHAP difference matrix visualization
c8_hyperparameter_xgboost.Rmd	Hyperparameter Tuning for XGBoost
c9_hyperparameter_ranger.Rmd	Hyperparameter Tuning for ranger
c10_rf_null.Rmd	Null distribution of RF prediction accuracy

---

📂 data

Multi-omics data (Genome, Microbiome, Metabolome, Phenotype) ready for analysis.

File	Description
SoyData_Drought2.RDS	Multi omics data - Drought
SoyData_Control2.RDS	Multi omics data - Control

---

📂 data/genome

Genome marker data (with different LD thresholds)

File	Description
genoMarker_LD0.001_SNP3078.RDS	SNP markers filtered at LD = 0.001
genoMarker_LD0.01_SNP10143.RDS	SNP markers filtered at LD = 0.01
genoMarker_LD0.1_SNP16419.RDS	SNP markers filtered at LD = 0.1
genoMarker_LD0.3_SNP34632.RDS	SNP markers filtered at LD = 0.3
genoMarker_LD0.001_SNP3078.csv	SNP markers filtered at LD = 0.001
genoMarker_LD0.01_SNP10143.csv	SNP markers filtered at LD = 0.01
genoMarker_LD0.1_SNP16419.csv	SNP markers filtered at LD = 0.1
genoMarker_LD0.3_SNP34632.csv	SNP markers filtered at LD = 0.3

---

📂 data/raw

Multi-Omics data raw files. Data include both control and drought conditions.

File	Description
metabolome.csv	Metabolome (265 features)
microbiome.csv	Microbiome (16457 features)
phenotype.csv	Phenotype (9 features)

---

For microbiome and metabolome, also refer to Dang et al. (2025). "I-SVVS: integrative stochastic variational variable selection to explore joint patterns of multi-omics microbiome data" https://doi.org/10.1093/bib/bbaf132

For full SNPs data, also refer to Kanegae et al. (2021). "Whole-genome sequence diversity and association analysis of 198 soybean accessiens in mini-core collections" https://doi.org/10.1093/dnares/dsaa032