Artificial intelligence is learning to find extraterrestrial life

Artificial intelligence (AI) systems are learning to do many tasks that are complex for people. Rare is the week in which we do not know its application or advances in some new field. Well, this Monday, an American team reveals that an AI system is doing quite well – with 90% reliability – in carrying out a key step in the search for life outside Earth: distinguishing whether the origin of samples is it biological or not (abiotic). Or what is the same, if these samples indicate that there are (or were) living organisms in that place.

Searching for traces of extraterrestrial life, past or present, is the great desire of scientists. This research is carried out within our solar system with robotic missions such as those of the Curiosity or Perseverance vehicles on Mars, and in much more distant worlds (exoplanets or extrasolar planets) through other techniques and telescopes that try to detect those worlds outside. of the Solar System biosignatures or biomarkers, that is, elements that may indicate signs of life, as we know it on Earth.

Among these biosignatures (elements, isotopes, molecules or phenomena that provide evidence that there has been or was life) are molecular oxygen, ozone or methane.. But the existence of one of these elements does not necessarily mean that there is life, that is, that it is a biomarker. For example, on Earth methane is produced by living beings such as bacteria or cows.. This gas has also been detected on Mars, but the origin could be both volcanism and biological processes.

In the same way, there are organic components that have been produced by biological activity (by living beings) or non-biological, and it is there, in the distinction between the two, where the artificial intelligence system that this Monday presents in the magazine is being trained. Proceedings of the National Academy of Sciences (PNAS), a team led by Jim Cleaves and Robert Hazen, from the Carnegie Institution for Science, in the USA.

As astrobiologist and mineralogy specialist Robert Hazen explains, they started from “the idea that the chemistry of life differs fundamentally from that of the inanimate world, that there are 'chemical rules of life' that influence the diversity and distribution of biomolecules. If we could deduce those rules, we could use them to guide our efforts to model the origins of life or to detect subtle signs of life on other worlds.”.

Their AI system was trained with molecular analysis data from 134 carbon-rich samples, both of biological and non-biological origin.. According to this study, AI was able to distinguish biotic samples from abiotic samples by detecting subtle differences in the molecular patterns obtained after analysis with instruments that separate and identify the components of a sample, and that determine the molecular weights of those components respectively.. Specifically, the AI was successful in correctly identifying the origin of samples of living organisms (such as modern shells, bones, teeth, insects, leaves or hair), remains of ancient life altered by geological processes (coal, carbon-rich fossils, oil or amber) and samples of non-biological origin.

According to Hazen, their analysis method “has the potential to revolutionize the search for extraterrestrial life and deepen our knowledge of the chemistry and origin of the first forms of life on Earth,” as they also want to use it to analyze ancient terrestrial rocks on those for which there is scientific debate. For example, 3.5 billion-year-old sediments found in Western Australia that some researchers say contain the oldest fossils of microbes, while others say they contain no traces of ancient life..

Rocks found in Australia that are 3.5 billion old CARNEGIE INST.

The American astrobiologist believes that this AI system could be incorporated into smart sensors that spacecraft and robotic vehicles would carry to search for signs of life before bringing the samples back to Earth..

Jorge Pla-García, researcher at the Center for Astrobiology (CAB/CSIC-INTA), not linked to the study published in PNAS, considers that it is “a very interesting investigation that could help astrobiologists in the future when it comes to determine if, indeed, any of the samples analyzed outside of Earth are really biomarkers”. This scientist, who is a member of the Spanish team that has supplied NASA with the weather stations carried by its Martian rovers and has signed numerous studies on the presence of methane on Mars, remembers “that only an organic compound that clearly and univocally comes from of biological activity. And this is not so easy to discern. On our own planet it is difficult to find and confirm signs of past life in rocks from the Early Earth (as the first phases of our planet are known). If doing this here at home is really complex, imagine doing it remotely on Mars, a planet far away from us at an average distance of 225 million km.”

To demonstrate that a sample is of biological origin, adds Pla-García, “not only must it be possible to demonstrate that life can create it, but it must also be ruled out that it was created by other processes.”. It is precisely at this point where AI plays a fundamental role according to this team of researchers and why the results seem very promising to me, although we also have to be cautious, since they talk about an accuracy of 90%, a fairly high value but not enough to unequivocally discern whether a compound comes from biological activity or not (in Astrophysics for example, to confirm that a compound is present in the atmosphere of an exoplanet, we need a precision of 99.977%)”, he points out.

“One of the big problems that the astrobiological community faces daily is the in situ analysis of samples, due to the limited performance of the instrumentation on board space missions compared to the very powerful and ambitious laboratories that we have on Earth.. “It is for this reason that to try to identify the origin of the Martian samples identified as being of high astrobiological interest according to the Perseverance rover, we need to bring them to Earth with the future Mars Sample Return (MSR) mission.”

However, as the Spanish scientist recalls, an independent review published last week concludes that this mission could cost more than 10 billion euros and suggests that NASA delay or replan it: “Perhaps, in the future, AI will help us study samples remotely without having to bring them home. This new research opens a new range of possibilities,” says Pla-García, who considers that AI is “a very powerful tool” that is already being used in his own research group to improve weather predictions on Mars, he says.