New AI tool can aid scientists in hunting for life on Mars

Kimberley Warren-Rhodes, a senior research scientist at the SETI Institute in Mountain View, California, and lead author of the paper, combined her background in statistical ecology with AI to help mission scientists "who are under a lot of pressure to find biosignatures."

An AI-based neural network and machine-learning algorithm helped search for life

In 2016, Warren-Rhodes’ group collected drone footage, geochemical analyses, and DNA sequences from an elevation of around 3,500 meters in the Chilean Andes, also the "proposed" Mars analog. The collected data set would be the same kind of information that researchers are acquiring on Mars with orbital satellites, drones, and rovers.

The team fed the data into an AI-based convolutional neural network (CNN) and a machine-learning algorithm. This then predicted where life was highly probable in the Atacama.

The AI helped the researchers reduce the search area by up to 97 percent and increase their likelihood of finding life by up to 88 percent.

"I'm very impressed and very happy to see this suite of work,” says Kennda Lynch, an astrobiologist at the Lunar and Planetary Institute in Houston, Texas, who studies biosignatures, told Nature. "It’s really cool that they can show some success with an AI to help predict where to go and look."

Further work is needed to be done. The new method will need to be verified across various ecosystems. as the Atacama is relatively simple when it comes to habitats and the types of lives found. According to Warren-Rhodes, the team's advance represents "an important advance in extraterrestrial research, in which biology has often lagged behind chemistry and geology."

"To have our team make one of these first steps towards reliably detecting biosignatures using AI is exciting," she said.

The study was published in Nature Astronomy.

Study Abstract:

In the search for biosignatures on Mars, there is an abundance of data from orbiters and rovers to characterize global and regional habitability, but much less information is available at the scales and resolutions of microbial habitats and biosignatures. Understanding whether the distribution of terrestrial biosignatures is characterized by recognizable and predictable patterns could yield signposts to optimize search efforts for life on other terrestrial planets. We advance an adaptable framework that couples statistical ecology with deep learning to recognize and predict biosignature patterns at nested spatial scales in a poly extreme terrestrial environment. Drone flight imagery connected simulated HiRISE data to ground surveys, spectroscopy and biosignature mapping to reveal predictable distributions linked to environmental factors. Artificial intelligence–machine learning models successfully identified geologic features with high probabilities for containing biosignatures at spatial scales relevant to rover-based astrobiology exploration. Targeted approaches augmented by deep learning delivered 56.9–87.5% probabilities of biosignature detection versus <10% for random searches and reduced the physical search space by 85–97%. Libraries of biosignature distributions, detection probabilities, predictive models and search roadmaps for many terrestrial environments will standardize analogue science research, enabling agnostic comparisons at all scales.