Artificial Intelligence Can Assess Signs of Life in Mars Samples

Researchers from the Carnegie Institution for Science, Purdue University, and Johns Hopkins University have developed a test that can quickly assess whether there are signs of life in materials. Although their model has only been tested on Earth samples so far, they hope that this test can be used to screen samples obtained by the Mars Curiosity rover. The test utilizes a method called Pyr-GC-El-MS, which combines pyrolysis gas chromatography and electron collision ionization mass spectrometry. This combination allows scientists to analyze tiny fragments of samples (weighing about 100 micrograms) to observe the composition of compounds in the samples, regardless of their molecular weight. This makes the test applicable to both biological (life-related) and non-biological (life-unrelated) samples. Traditionally, electron mass spectrometry results are inputted into a database and matched with known compounds. However, when the goal is to search for life on other planets, the use of known compounds is limited; after all, extraterrestrial life may involve compounds that we have not observed on Earth. Based on this reasoning, the researchers' test does not look for specific molecules; instead, it searches for small differences in the distribution and diversity of molecules and compounds. The logic behind this idea is that the patterns within Earth biological samples (rather than the molecules that make them up) are similar to patterns on other planets. Determining and comparing these patterns requires a significant amount of work, making it an ideal task for artificial intelligence. The researchers used Pyr-GC-EI-MS technology to analyze 134 known carbon samples, including teeth, coal, shells, oil shale, carbonaceous meteorites, and pure synthetic chemicals. They then trained a machine learning model to map out the molecular patterns found in each sample. To test the model, the researchers introduced new samples. The model was able to distinguish between biological and non-biological samples with 90% accuracy, and even began categorizing biological samples as "living" or "fossil." In a paper published in the Proceedings of the National Academy of Sciences (PNAS), the researchers wrote, "Biological systems on other planets may not produce organic molecules that are identical or even roughly similar to those found in modern Earth biology. Instead, we propose that even alien biochemistry systems that differ significantly from Earth's biochemistry would still exhibit molecular frequency distributions distinct from those of abiotic synthesis processes... We suggest that our technique can detect and quantify these differences in the composition of biological and non-biological molecules." The model has already been used in space missions, and it is hoped that it will help future explorers in detecting signs of extraterrestrial life.