Update:
In the reporting period, a safe and successful harvest was completed in the fall of 2023. Post-harvest, seed processing, and trait data acquisition were also completed in late fall and winter. Selection decisions on advancements were made. Additionally, plant material was sent to the winter nursery to get the second crop cycle. Preparations for 2024 planting were initiated, and at the time of reporting, packaging for 2024 was in full swing.
Student updates on their research projects:
Matt Carroll graduated with a Ph.D. He is a research scientist in the Iowa Soybean Association's analytics team. Matt and co-investigators submitted a paper for peer review from a project that leverages soil mapping and machine learning to improve spatial adjustments in plant breeding trials. Spatial adjustments are routinely made in breeding trials to improve the estimate of plot seed yield. Moving mean and P-Spline are examples of spatial adjustment methods used in plant breeding trials to deal with field heterogeneity. Within trial spatial variability primarily comes from soil feature gradients. Therefore, information on soil nutrients and important soil factors must be included in breeding trials. We analyzed plant breeding progeny row and preliminary yield trial data of ISU's soybean breeding program across three years consisting of >43K plots. We compared several spatial adjustment methods: unadjusted (as a control), moving means adjustment, P-spline adjustment, and XGBoost. We established the usefulness of spatial adjustments at both progeny row and preliminary yield trial stages of field testing. We also provide ways to utilize interpretability insights of soil features in spatial adjustments. These results will empower breeders in public institutions and private companies to refine selection criteria to make more accurate selections and include soil variables to select for macro- and micro-nutrient stress tolerance in their breeding efforts.
Our team members were also involved in a project aimed at yield prediction. Soybean yield prediction is a challenging problem in plant breeding that is often simultaneously affected by many factors. Hyperspectral reflectance data from plants and soil data provide breeders with useful information about soybean plant health, and using these different types of data to predict yield is an active area of research. Furthermore, breeding programs encounter challenges such as data imbalance and external factors like genotype variability across different environments. These present significant hurdles in developing yield prediction models for large-scale breeding programs. In an interdisciplinary project, we predicted yield using hyperspectral reflectance and soil data to understand what scenarios offer the best chances to predict yield accurately. We reported an improved pipeline for yield prediction using hyperspectral reflectance data that performs well for large-scale breeding programs.
Matheus Krause, a Ph.D. student with me and Dr. Beavis, graduated with a Ph.D. from ISU and now works for Corteva. Matheus's published paper that examined models to estimate genetic gain of soybean seed yield from annual multi-environment field trials. In this study, we recommended multiple selected models to obtain a range of reasonable estimates. These results will be helpful to plant breeders.
Sarah Jones, Ph.D. student in the team, processed drone imagery from 3 years of drought stress screening of 450 PI lines and collaborated with an engineering colleague (from Dr. Sarkar's group) to develop a ML pipeline for drought severity classification and early detection.
Liza Van der Laan is preparing her thesis chapter on the screening a diverse panel of 450 soybean accession for heat stress tolerance. This project aims to identify heat tolerant accessions that will be useful for breeding for abiotic stress tolerance, as well as to understand the genetic mechanisms controlling heat stress tolerance. The protocol for screening heat stress tolerance and other abiotic stressors of drought and flood was expanded in our program this past year with work on developing a field based contained stress platform.
Sam Blair has continued researching drone-based relative maturity and ground robot yield estimations. Large-scale data collection using RGB imagery for the drone-based relative maturity estimation project was completed in 2023. ML models were trained to make automatic maturity ratings. This model has shown promising results, and we aim to deploy it in the 2024 field season. The results will be useful to other breeding and field research programs in the public and private sectors.
Based on multi-state testing for yield, agronomics, quality and yield protection traits, foundation seed increase was done for previous releases. Two new invention disclosures were made to ISURF during the reporting period. One of the new varieties is a conventional non-GM line with high yield and suitable for production in central and southern IA. The other variety is combines good yield and higher protein and yellow hilum for soybean food industry. This new variety is suitable for IA production regions. This variety provides farmers access to a higher protein line with protein levels of >42% on dry weight basis compared to generic beans that have ~37%.
We continued our partnership with the Iowa Soybean Association with funded projects, AIIRA, COALESCE, and SIRAC, that USDA and NSF supported. These unique linkages provide farmer-motivated solutions that enhance breeding and crop production through the use of machine learning and robotics.
In 2023, we entered ISU experimental lines in the Uniform Soybean tests, which are grown in locations across several mid-western states. Several candidate varieties performed well. Before commercialization, we will complete the testing of our most advanced and promising lines in multiple states.