Longitudinal Modeling of Maize Gas Exchange for Optimizing Water Use Efficiency

Stomatal conductance (gsw), a measure of gas diffusion through stomata, has significant implications for water-use efficiency (WUE) in plants. The slow anion channel 1 (slac1) gene is a key candidate for engineering drought resilience through targeted programming of slac1. This study investigates how genetic variation at the slac1 locus, combined with diverse genetic backgrounds, affects diurnal patterns of stomatal conductance in maize (Zea mays L.). Utilizing an extreme UniformMu stomatal mutant (slac1-2, mu1037824) as a prototype Programmable Plant System (PPS), we crossed this mutant and its wild-type counterpart with eight maize inbred lines with distinct expression levels of slac1, generating 16 hybrid genotypes. Stomatal conductance was measured using LI-600 porometers across replicated field trials during the 2023 and 2024 growing seasons. Data were analyzed using a two-stage Gaussian and Linear Mixed Effects model to elucidate the factors contributing to variation in gsw. Our analysis examined genotype, environmental, and mutant allele effects, revealing significant genotype-specific differences in gsw. Further investigation through Random Forest analysis, identified major environmental factors responsible for variation not explained by genetic differences. This work demonstrates how variation at the slac1 locus can effectively shift peak stomatal conductance to earlier, cooler periods of the day, reducing mid-day water loss. Additionally, this research underscores the substantial impact environmental conditions have on complex traits like drought tolerance, illustrating the challenges in isolating genomic effects. We characterized some aspects of this underlying complexity and separately observed instances where gsw was favorably altered during periods of low VPD, showcasing the potential for programming WUE through slac1 allelic variation in maize.

Over the summer I had the opportunity to work in both Micheal Gore and Kelly Robbins’ labs as part of a co-mentored CROPPS project, funded by the NSF. I came to this opportunity with limited experience in quantitative plant genomics but with the strong support network at Cornell, CROPPS, and BTI, I was able to effectively leverage my past experience at the University at Buffalo with statistics and programming, in assembling the genome of the carnivorous purple pitcher plant and studying its evolutionary features with Victor Albert, as well as past experience using open source machine learning architectures to interrogate sequence data to find oxidative legions with Jaroslaw Zola. I built on these experiences by diving more into statistical modeling and traditional methods of ML like Random Forest regression. Over my time here I conducted a split 2-stage analysis of a diverse population of maize carrying mutated and wild-type slac1 alleles. I first fit the data to a Gaussian model, then forward-fit a Linear Mixed Effect model of the data to determine linear relationships to what factors, weather environmental, genotypic, or allele, is causing the most variation in the data and thus the phenotype. I used Random Forest Regression to determine which of the environmental factors were most important in affecting the phenotype. My responsibilities spanned the entire computational pipeline, including data preprocessing and dataset construction, to running the 2-stage gaussian/LME model, to then running random forest regression in parallel on Cornell’s BIO HPC cluster and validating the results. I sharpened my skills with field and experimental design, as well as how to think about statistically modeling real world phenomena with the intention to predict and modify such phenomena. I learned how to select appropriate modeling methods based on specific criteria, adopting an agnostic approach toward method selection as ultimately, the goal is accuracy, regardless of what modeling approach is employed. Furthermore, I learned how to layer multiple models together to derive more robust and insightful results. After this program I will be finishing my last year of undergraduate studies at the University at Buffalo and plan to begin graduate studies after to continue working on real world foundation modeling. I am primarily looking for PhD positions but remain open to other opportunities. My main research interests now revolve around understanding foundational functions of genomic sequences and modeling how sequence alterations affect phenotype, with additional interest in applying these insights to synthetic biology research, enhancing food security and human health, and upholding our national interests.