Brett McKinney, Ph.D., Warren Foundation Chair Professor of Computer Science and Mathematics, and dean fellow at The University of Tulsa, is at the forefront of interdisciplinary research, blending expertise in machine learning (ML), bioinformatics, and space science. A journey that began as a UTulsa student has since sprouted into a career garnering support from NASA to focus on detecting biosignatures from ocean world analogue samples. McKinney was recently awarded the Established Program to Stimulate Competitive Research (EPSCoR) grant from NASA, the sole national EPSCoR grant awarded in Oklahoma for 2024. UTulsa will receive $750,000 over three years in grant funding.
Simulating ocean worlds with NASA
McKinney’s NASA-funded astrobiology project began with a collaboration between former UTulsa geoscience faculty member and NASA scientist, Bethany Theiling, who sought ML expertise to analyze mass spectrometry data. This partnership supported UTulsa geosciences graduate student Lily Clough, who is now a key contributor to the new NASA-funded project. These collaborations have fostered additional NASA internship opportunities for UTulsa students.
The team’s research focuses on detecting biosignatures in mass spectrometry data, aiming to differentiate biological from abiotic samples. “It’s a bioinformatics problem in the context of space science,” McKinney explains.
By integrating geochemistry, biochemistry, and ML, the team is tackling challenges in identifying extraterrestrial life on icy moons like Europa and Enceladus. This process includes simulating extreme ocean world environments to examine how extremophile microbes influence geochemistry under non-Earth conditions. “This project demands expertise from nearly every scientific and computational discipline,” McKinney notes, emphasizing the role of collaboration in driving innovation.
McKinney’s UTulsa colleagues Hema Ramsurn, Ph.D., A. Paul Buthod Endowed Chair in Chemical Engineering and associate professor of chemical engineering, and Mohamed Fakhr, Ph.D., professor of biological sciences, are involved, along with researchers from the University of Oklahoma, Oklahoma State University, and Northeastern Oklahoma Agricultural and Mechanical. This larger team is generating data to create ML models rooted in biogeochemistry.
Overcoming challenges for space exploration
Applying ML to ocean world analog samples presents significant challenges, particularly concerning the trust and explainability of models. “Machine learning models are often perceived as black boxes,” explains McKinney. “This makes it difficult to understand the rationale behind their predictions.” This opacity can lead to mistrust, especially when predictions dictate critical actions, such as detecting extraterrestrial life.
Grounding the ML models in domain-specific knowledge, such as geochemistry, enhances the reliability of the black box predictions. By making models more interpretable, scientists can better assess the validity of predictions, thereby increasing confidence in the use of ML for analyzing ocean world analogue samples.
In July 2024, McKinney, Fakhr, and Clough collected soil and brine samples at the Great Salt Plains that will be used in comparison with experiments for complex planetary analogues such as Mars, Enceladus, and Europa. In addition, UTulsa undergraduate interns tested a prototype of an environmental sensing tool on the blue box. They demonstrated automated data collection, geotagging, environmental anomaly detection, and ML predictions for soil samples.
A stellar future
McKinney’s long-term goal is to see his geochemistry-informed ML algorithms deployed on NASA missions. Beyond space exploration, the research could yield insights into Earth’s oceans and evolutionary processes. The team is also exploring optimal control methods to understand microbial metabolism and quantum computing’s potential in advancing ML.
Another project is on the horizon involving optimal control methods. This would mean new research into how microbial life uses different molecules and isotopes to optimize metabolism, ultimately leading to better understanding of the rules of life and evolution. In the biomedical domain, McKinney’s group hopes to use gene expression data to create ML models of a person’s biological age and how it affects health.
Through pioneering research, McKinney exemplifies how UTulsa’s collaborative environment and interdisciplinary focus are shaping the future of science and technology. Strong collaboration with geosciences faculty and students, whose expertise lies in understanding the planetary process, will be key to advancing ML and biosignature detection. This research is inspiring the next generation of innovators at UTulsa and beyond.
