Timezone: »

Inferring molecular complexity from mass spectrometry data using machine learning
Timothy Gebhard · Aaron C. Bell · Jian Gong · Jaden J. A. Hastings · George Fricke · Nathalie Cabrol · Scott Sandford · Michael Phillips · Kimberley Warren-Rhodes · Atilim Gunes Baydin

Molecular complexity has been proposed as a potential agnostic biosignature — in other words: a way to search for signs of life beyond Earth without relying on “life as we know it.” More than one way to compute molecular complexity has been proposed, so comparing their performance in evaluating experimental data collected in situ, such as on board a probe or rover exploring another planet, is imperative. Here, we report the results of an attempt to deploy multiple machine learning (ML) techniques to predict molecular complexity scores directly from mass spectrometry data. Our initial results are encouraging and may provide fruitful guidance toward determining which complexity measures are best suited for use with experimental data. Beyond the search for signs of life, this approach is likewise valuable for studying the chemical composition of samples to assist decisions made by the rover or probe, and may thus contribute toward supporting the need for greater autonomy.

Author Information

Timothy Gebhard (Max Planck Institute for Intelligent Systems, Tübingen)
Aaron C. Bell (Insight Edge Inc.)
Aaron C. Bell

Aaron originally studied biology, and then moved into the field of astronomy as his interests evolved. While researching the Milky Way, and the kinds of dust and molecules floating around between the stars, he took a strong interest in the analysis of large astronomical “all-sky” surveys, taken by cutting-edge space telescopes. After finishing graduate school, Aaron continued astronomical research for a time, and joined the short-term research accelerator program, NASA Frontier Development Lab. Loving this type of fast-paced, collaborative atmosphere, he decided to move to industry, first working at a Tokyo-based AI solutions startup as an engineer, and finally joining Insight Edge as Data Scientist in May 2021. He has experience with image analysis, machine learning, data processing, and anomaly detection. Aaron continues to be interested in space, computing, and solving social problems through a balance technology, curiosity, and empathy. Aaron returned to the Frontier Development Lab for its 2022 iteration, as a researcher again with the Astrobiology Challenge Team, implementing machine learning in the search for agnostic biosignatures in space. Ph.D. in Astronomy, University of Tokyo, Graduate School of Science.

Jian Gong (Massachusetts Institute of Technology)
Jaden J. A. Hastings (XO.LABS)
George Fricke (University of New Mexico)
Nathalie Cabrol (SETI Institute)
Scott Sandford (NASA Ames Research Center)
Michael Phillips (Johns Hopkins Applied Physics Laboratory)

As a planetary geologist, I use remote sensing data and mathematical models to understand planetary bodies, their composition, geology and geochemistry, formation histories, and potential to host life. My research is focused on planetary surface geology and processes and applications of AI/ML to astrobiological exploration. The data I use in my research are hyper- and multispectral reflectance and thermal emission spectra, topographic data, and various other satellite- and small Unmanned Aerial System (sUAS)-based products. To augment my remote sensing approach, I conduct field work, employ numerical, analytical, and machine learning models, and perform laboratory and field spectroscopy. My research is highly collaborative, and I work with leading scientists in their field across many institutions, including at The University of Tennessee, Knoxville (UTK) and the SETI Institute (SI) where I am a Research Affiliate at both institutions, and at my home institution, The Johns Hopkins University Applied Physics Laboratory (APL).

Kimberley Warren-Rhodes (SETI Institute)
Atilim Gunes Baydin (University of Oxford)

More from the Same Authors