Particle.news
Download on the App Store

AI-Enabled Analysis Detects 3.3-Billion-Year-Old Biosignatures and Dates Photosynthesis to 2.5 Billion Years

Researchers paired pyrolysis–GC–MS with supervised machine learning to read chemical echoes of life with over 90% accuracy.

Overview

  • The peer-reviewed study, led by the Carnegie Institution for Science, was published in the Proceedings of the National Academy of Sciences on November 17, 2025.
  • The workflow analyzed more than 400 reference samples—including modern organisms, billion-year-old fossils, and meteorites—to train a supervised model that distinguishes biotic from abiotic chemistry.
  • Chemical fragments in Archean sediments older than 3.3 billion years were classified as life-related, with key samples reported from South Africa.
  • Signals consistent with oxygen-producing photosynthesis were detected in rocks about 2.52 billion years old, pushing prior evidence back by roughly 800 million years and informing models of Earth’s oxygenation.
  • The method roughly doubles the reliable molecular biosignature window from about 1.7 to over 3.3 billion years and offers a calibrated approach for future analysis of extraterrestrial samples, subject to stringent validation and contamination controls.