Machine learning model for predicting banana crop productivity from soil microbiome data

Overview

Banana farmers face significant challenges due to variable yields driven by environmental and soil factors, making production planning difficult. This research initiative from Universidad Tecnológica de Pereira aims to develop a predictive model that links soil microbiome composition to banana crop productivity. By applying metagenomic sequencing and machine learning techniques, the project seeks to uncover microbial patterns that correlate with higher yields, ultimately providing farmers and agricultural planners with a data-driven decision-making tool.

The solution combines soil microbiome characterization (bacteria and fungi) with physical-chemical soil analysis to build an expert system capable of predicting productivity based on microbial consortium abundance. This approach addresses a critical gap in precision agriculture for banana cultivation, one of the most important crops in Colombia's coffee-growing region.

Technical specifications

Research approach:

• Metagenomic analysis of soil and rhizosphere samples using 16S rDNA and ITS (Internal Transcribed Spacer) region amplification and sequencing
• Microbiota characterization under different cultivation schemes, including monoculture and banana-coffee associated cropping systems
• Physical-chemical soil characterization to complement biological data
• Bioinformatic analysis and Metagenomic Wide Association Studies (MWAS) to identify microbial consortia linked to productivity
• Machine learning model construction for productivity prediction based on microbial abundance patterns
• Development of an expert system for metagenomic-data-driven productivity forecasting
• Isolation and multiplication of beneficial soil microbial communities using alternative biological models

Key scientific basis:

The rhizosphere harbors plant growth-promoting rhizobacteria that influence crop performance through phytohormone secretion, biological nitrogen fixation, and phosphorus solubilization. Understanding these microbial communities enables identification of beneficial consortia for soil health and yield optimization.

Technology readiness level

The project builds on prior characterization work by the research group, which has already completed metagenomic profiling of microbial communities in banana crops under different cultivation schemes. The proposed 24-month study will advance the work through soil sampling, DNA extraction and sequencing, bioinformatic analysis, model construction, and expert system development. The research operates under a Framework Agreement for Access to Genetic Resources with the Colombian Ministry of Environment, ensuring regulatory compliance for genetic resource use.