Q&A 24 How do you use mikropml in R for microbiome machine learning?
24.1 Explanation
mikropml is a microbiome-focused R package by Pat Schloss designed for:
- End-to-end modeling workflows
- Built-in cross-validation and hyperparameter tuning
- Transparency in model reporting and evaluation
It simplifies the process of building, tuning, and interpreting microbiome ML models.
This Q&A introduces a basic pipeline using mikropml and prepares the OTU and metadata files as expected.
24.2 R Code
# 📦 Ensure mikropml is installed
if (!requireNamespace("mikropml", quietly = TRUE)) {
if (!requireNamespace("remotes", quietly = TRUE)) install.packages("remotes")
remotes::install_github("SchlossLab/mikropml")
}
library(mikropml)
library(tidyverse)
# Load OTU table and metadata
otu <- read.delim("data/otu_table_filtered.tsv", row.names = 1)
meta <- read.delim("data/sample_metadata.tsv")
# Transpose OTU so samples are rows
otu_t <- t(otu)
otu_df <- as.data.frame(otu_t)
otu_df$sample_id <- rownames(otu_t)
# Merge with metadata
data <- inner_join(otu_df, meta, by = "sample_id")
# Run mikropml using run_ml()
set.seed(42)
fit <- run_ml(
dataset = data,
outcome_colname = "group",
method = "rf", # Choose from rf, svm, glmnet, xgb
seed = 42
)
# View model summary
summary(fit) Length Class Mode
trained_model 21 train list
test_data 55 data.frame list
performance 17 tbl_df list
feature_importance 1 -none- characterNULL