Machine learning insights into environmental and biological triggers of fumonisin outbreaks in Texas corn

Contamination of corn with fumonisin mycotoxins (FUM) caused by the fungus Fusarium verticillioides is a recurrent issue in Texas corn production that results in significant economic losses for corn producers. Both prediction and management of FUM outbreaks are challenging due to a lack of information regarding region-specific environmental and biological drivers of contamination. We analyzed 683 historical FUM contamination events for 12 years that were categorized as low (<5 mg kg −1 ; <5 ppm) or high (≥5 mg kg −1 ; ≥5 ppm) and evaluated geo-spatial, environmental and biological drivers of outbreaks using a machine learning approach. We applied Ratkowsky-based growth and toxin production models for five genetically diverse F. verticillioides strains and found strain-specific predicted growth and Fumonisin Risk Index (FRI) profiles. Using FRI (strain-specific) or weather-only feature sets, each combined with soil biogeochemistry properties, we developed Random Forest (RF) and Neural Network (NNET) models. Historical balanced accuracy averaged ~74% for RF and ~70% for NNET models. Feature importance revealed that RF models consistently ranked barometric pressure and latitude among top predictors, while NNET models emphasized soil moisture and FRI. Among the overlapping drivers between RF and NNET were FRIs at the beginning of February and mid-October. Critical influential weeks for soil moisture (May–August), barometric pressure (July–October), and FRI (August–November) overlapped with corn developmental stages susceptible to F. verticillioides infection and FUM accumulation in a region-dependent manner due to differences in planting dates across the state. Soil biogeochemical properties such as organic matter, electrical conductivity, and clay content were also influential in models’ predictions indicating region-specific relationships with historical outbreaks. This study highlights the combined roles of weather patterns, fungal strain diversity, soil properties, and crop phenology in shaping FUM outbreak risk across Texas and underscores the importance of integrating strain-specific biological responses and regionally relevant environmental factors into predictive modeling frameworks for improved mycotoxin risk assessment. In addition, results suggest that region specific management of soil physico-chemical properties could reduce FUM risk, offering insights into new mitigation strategies for on-farm prevention of mycotoxin contamination and the potential to develop a novel early-warning system as well as mitigation tools for FUM outbreaks in Texas.

Authors

Institutions

Publication Details

Journal
Frontiers in Microbiology
Published
2026-11-16
DOI
https://doi.org/10.3389/fmicb.2026.1993463
Primary Topic
Mycotoxins in Agriculture and Food
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Machine learning insights into environmental and biological triggers of fumonisin outbreaks in Texas corn

Robert H. Proctor, Lina Castano‐Duque, Martha Marie Vaughan, Ronnie W. Schnell et al.
2 links
Frontiers in Microbiology
Mycotoxins in Agriculture and Food
article

Machine learning insights into environmental and biological triggers of fumonisin outbreaks in Texas corn

Robert H. Proctor, Lina Castano‐Duque, Martha Marie Vaughan, Ronnie W. Schnell, Hillary L. Mehl, Brian M. Mack, Mark Busman, Joshua Blackstock, Joseph Opoku, H. Edwin Winzeler, Thomas Isakeit, Jianzhong Su, Phillip Ray Owens, Angela Avila
article en

Abstract

Contamination of corn with fumonisin mycotoxins (FUM) caused by the fungus Fusarium verticillioides is a recurrent issue in Texas corn production that results in significant economic losses for corn producers. Both prediction and management of FUM outbreaks are challenging due to a lack of information regarding region-specific environmental and biological drivers of contamination. We analyzed 683 historical FUM contamination events for 12 years that were categorized as low (<5 mg kg −1 ; <5 ppm) or high (≥5 mg kg −1 ; ≥5 ppm) and evaluated geo-spatial, environmental and biological drivers of outbreaks using a machine learning approach. We applied Ratkowsky-based growth and toxin production models for five genetically diverse F. verticillioides strains and found strain-specific predicted growth and Fumonisin Risk Index (FRI) profiles. Using FRI (strain-specific) or weather-only feature sets, each combined with soil biogeochemistry properties, we developed Random Forest (RF) and Neural Network (NNET) models. Historical balanced accuracy averaged ~74% for RF and ~70% for NNET models. Feature importance revealed that RF models consistently ranked barometric pressure and latitude among top predictors, while NNET models emphasized soil moisture and FRI. Among the overlapping drivers between RF and NNET were FRIs at the beginning of February and mid-October. Critical influential weeks for soil moisture (May–August), barometric pressure (July–October), and FRI (August–November) overlapped with corn developmental stages susceptible to F. verticillioides infection and FUM accumulation in a region-dependent manner due to differences in planting dates across the state. Soil biogeochemical properties such as organic matter, electrical conductivity, and clay content were also influential in models’ predictions indicating region-specific relationships with historical outbreaks. This study highlights the combined roles of weather patterns, fungal strain diversity, soil properties, and crop phenology in shaping FUM outbreak risk across Texas and underscores the importance of integrating strain-specific biological responses and regionally relevant environmental factors into predictive modeling frameworks for improved mycotoxin risk assessment. In addition, results suggest that region specific management of soil physico-chemical properties could reduce FUM risk, offering insights into new mitigation strategies for on-farm prevention of mycotoxin contamination and the potential to develop a novel early-warning system as well as mitigation tools for FUM outbreaks in Texas.

Frontiers in MicrobiologyVol. 17
Agricultural Research Service (US), United States Department of Agriculture (US), National Center for Agricultural Utilization Research (US), University of Arkansas System (US), Southern Regional Research Center (US), Dale Bumpers Small Farms Research Center (US), U.S. Arid Land Agricultural Research Center (US), University of Arkansas at Fayetteville (US), Texas A&M University (US)
Zero hunger
Mycotoxins in Agriculture and Food
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon