Madrid, 13 years old (Europe Press)
A new machine learning technology evaluates complex organic mixtures with a mass spectrometer, for reliable confirmation of potential biomarkers in other worlds.
There are currently many ways that scientists are searching for extraterrestrial life. This includes listening to radio signals from advanced civilizations in deep space, looking for subtle differences in the atmospheric composition of planets around other stars, and trying to measure them directly in soil and ice samples that they can collect using spacecraft in our solar system. The latter category allows them to bring their more advanced chemical analytical tools straight to samples, and perhaps even bring some samples back to Earth, where they can be carefully examined.
Missions like the NASA Perseverance rover will search for life on Mars this year. NASA’s Europa Clipper program, set to launch in 2024, will attempt to sample ice ejected from Jupiter’s moon Europa, and the Dragonfly mission will attempt to land an “octacopter” on the Saturn Titan in 2027. All of these missions will attempt to answer the question of whether we are Alone.
Mass spectrometry (MS) is an essential technology that scientists will rely on in spacecraft-based searches for extraterrestrial life. MS has the advantage of being able to simultaneously measure a large number of compounds in samples and thus provide a kind of “fingerprint” of the sample composition. However, interpreting these fingerprints can be difficult.
As far as scientists can tell, all life on Earth is based on the same highly coordinated molecular principles, giving scientists confidence that all life on Earth is derived from a common ancient earthly ancestor.
However, in simulating the primitive processes that scientists believe may have contributed to the genesis of life on Earth, many similar but slightly different versions of the particular molecules used by life on Earth are often discovered. In addition, naturally occurring chemical processes can also produce many of the building blocks of biological molecules.
With no known sample of extraterrestrial life yet, this leaves scientists with a conceptual paradox: Did terrestrial life make some arbitrary decisions early in evolution that were banned and thus life could be constructed otherwise? The whole life? Everywhere it has to be exactly the same as it is on Earth? How can we know that the discovery of a specific type of molecule indicates whether or not it has been produced by extraterrestrial life?
Scientists have long been concerned that biases in the way we think life should be discoverable, which largely depends on what life looks like on Earth today, could cause our detection methods to fail. In fact, Viking 2 returned strange results from Mars in 1976. Some of its tests gave signs deemed positive for life, but the MS measurements did not provide evidence of life as we know it.
The latest MS data from NASA’s Curiosity rover indicates the presence of organic compounds on Mars, but it does not yet provide evidence of life. A related problem afflicted scientists trying to discover the oldest evidence of life on Earth: How can we know whether the signals detected in ancient terrestrial samples are from the original living organisms preserved in those samples or are they derived from pollution from living organisms currently spreading our planet?
Scientists from the Institute of Earth and Life Sciences at Tokyo Institute of Technology in Japan and The National MagLab in the USA decided to tackle this problem using a combined experimental and computational approach. Machine learning. They publish the results in the scientific journal Life.
Using ultra-fine MS technology (a technique known as Fourier transform cyclotron resonance spectrometry (or FT-ICR MS)) they measured the mass spectra of a variety of complex organic mixtures, including those derived from in vitro biological samples, and mixtures Organics found in meteorites, laboratory-grown microorganisms, and crude oil derived from long-lived organisms on Earth provide an example of how the “fingerprint” of known organisms has changed over geological time.
Each of these samples contained tens of thousands of separate molecular compounds, providing a wide range of MS spectra that can be compared and classified.
In contrast to methods that use the precision of MS measurements to uniquely identify each vertex with a specific molecule in a complex organic mixture, the researchers pooled their data and looked at the overall statistics and signal distribution. Complex organic mixtures, such as those derived from living organisms, oil and biological samples, present very different ‘fingerprints’ when presented in this way. It is more difficult for a human to detect such patterns than to discover the presence or absence of individual types of molecules.
The researchers fed their raw data into a computer’s machine learning algorithm and surprisingly found that the algorithms could precisely classify samples as live or nonliving with 95% accuracy. Importantly, they did so after greatly simplifying the raw data, making it plausible that low-fidelity, spacecraft-based instruments often had low power, and could obtain data of sufficient precision to allow the accuracy of the biological classification obtained by the team.
The reasons behind this classification accuracy remain unexplored, but the team indicates that this is due to the ways in which biological processes, which modify organic compounds differently from biological processes, relate to biological processes. The processes that allow life to propagate. Living processes have to make copies of themselves, whereas biological processes have no internal process to control them.