AI for Predictive Crop Protection with Microbial Data Science

To train participants to build predictive and explainable AI models using soil microbiome datasets for disease/pest risk forecasting and decision-support.

• Build literacy in soil microbiome datasets and prediction targets.
• Preprocess microbiome + metadata for ML-ready modeling.
• Engineer meaningful microbial features (diversity, indicator taxa, transforms).
• Train, compare, and evaluate disease/pest risk prediction models.
• Apply explainable AI to identify key microbial drivers of risk.
• Create a reusable pipeline and a one-page decision-support report.

📅 Day 1 — Soil Microbiome Data Literacy for AI

• Soil microbiome intelligence: crop protection objectives and prediction targets
• Microbiome datasets: ASV/OTU tables, taxonomy, soil/crop metadata
• Preprocessing essentials: cleaning, normalization, feature-ready formatting
• Hands-on (Colab): Import a microbiome dataset + metadata, preprocess, and build a labeled “risk prediction” table

📅 Day 2 — Predictive Modeling with Microbial Features

• Feature engineering: compositional transforms, diversity features, indicator taxa
• Model development: classification workflow, imbalance handling, robust splitting
• Evaluation: cross-validation, ROC/PR metrics, performance interpretation
• Hands-on (Colab): Train and compare two models for disease/pest risk prediction and generate risk scores

📅 Day 3 — Explainable AI and Decision-Support Outputs

• Interpretability: microbial drivers of risk, stability checks, responsible biomarker insights
• Translating predictions to actions: intervention pathways and management recommendations
• Reusable pipeline: data → prediction → explanation → report
• Hands-on (Colab): Generate an explainable one-page report (risk + key drivers + recommendations) from the model in a notebook

• Students, researchers, and faculty in Agriculture, Microbiology, Biotechnology, Environmental Science, or related fields.
• Data science/AI learners and professionals interested in agritech and biological datasets.
• Basic familiarity with biology/microbiomes and comfort with spreadsheets; prior coding in Python is helpful but not mandatory (Colab guidance provided).

• Prepare microbiome datasets for ML (cleaning, normalization, labeling).
• Engineer features and handle imbalance for robust classification.
• Train and evaluate models using ROC/PR metrics and cross-validation.
• Generate risk scores and interpret microbial drivers using explainable AI.
• Produce a notebook-based one-page report with risk, key drivers, and recommendations.