Lightweight AI Tool AMPLiT Unearths Potent Antimicrobial Peptides from Millennia-Old Coprolites
Today, we share important research, soon to be published in Nature Communications, titled "Identification of antimicrobial peptides from ancient gut microbiomes." This study, by developing a lightweight artificial intelligence tool named AMPLiT, systematically analyzed metagenomic data from seven ancient human fecal fossils (coprolites) dating back 1000-2000 years. It successfully uncovered 160 potential antimicrobial peptide (AMP) candidates. Experimental validation showed that among the 40 peptides successfully synthesized, 36 (90%) exhibited significant antibacterial activity in vitro. Remarkably, approximately two-thirds of the active antimicrobial peptides originated from Segatella copri(formerly Prevotella copri), a dominant symbiotic bacterium in ancient guts whose prevalence has significantly declined in modern populations. These AMPs from ancient microbiomes not only possess membrane-disrupting mechanisms, low cytotoxicity, and low hemolysis risk but also demonstrated efficacy comparable to traditional antibiotics (vancomycin, polymyxin B) in a mouse wound infection model. This work reveals the immense potential of ancient gut microbiomes as a treasure trove for novel antimicrobial peptides, offering a new solution to combat the antibiotic resistance crisis.
01 Research Background
Antibiotic resistance has become a global health crisis, directly causing approximately 4.5 million deaths in 2019, urgently necessitating the development of new antimicrobials. The human gut microbiome is a vast reservoir of bioactive molecules, among which antimicrobial peptides are of particular interest due to their unique mechanisms (e.g., membrane disruption, immunomodulation) and low propensity for resistance. However, current research predominantly focuses on modern human microbiomes, which have co-evolved long-term with pathogens, potentially allowing pathogens to develop corresponding resistance mechanisms. Ancient gut microbiomes, preserved in samples like coprolites and dental calculus, act as "genetic time capsules" that record the evolutionary history of microbes under pre-industrial diets, natural antimicrobial exposure, and virtually no antibiotic pressure. These ancient microbial communities differ significantly in composition from modern microbiomes, with one notable feature being the dominance of Segatella copri. The abundance and prevalence of this bacterium have drastically declined in modern populations, especially those with Westernized lifestyles. Given their unique composition and the rapid evolutionary rate of microbial strains, ancient microbial ecosystems may harbor antimicrobial peptides with targets distinct from those in modern pathogens. These peptides evolved in battles against ancient pathogens lacking modern resistance mechanisms, and their molecular targets might remain vulnerable in contemporary pathogens, thus holding immense therapeutic potential.
02 Innovative Highlights
Development of Lightweight AI Tool AMPLiT for Efficient and Accurate AMP Mining:
Addressing the issue that existing computational methods for accurately identifying antimicrobial peptides from metagenomic data require substantial computational resources, the research team redesigned the architecture of their previous tool, AMPidentifier. By introducing three innovative computational modules, they achieved comparable predictive performance (AUPRC: 0.9486 ± 0.0003) while reducing training time by ~80% (requiring only 3200 ± 53 seconds) and enabling operation on consumer-grade hardware (Intel i7 CPU), significantly lowering the barrier for AMP discovery.
First Systematic Mining of Ancient Coprolite Metagenomes, Revealing S. copri as a Key Source of Novel AMPs:
This study is the first to computationally mine ancient human coprolite metagenomes, successfully identifying 160 candidate AMPs. The most critical finding is that approximately two-thirds of the active AMPs originated from S. copri. This directly links a declining ancient symbiotic bacterium in modern populations to the discovery of novel AMPs, providing a fresh perspective on understanding its historical ecological role and its potential role in modern health.
Construction of a Multi-Level, Rigorous Experimental Validation System from In VitroActivity to In VivoEfficacy:
The research extended beyond computational prediction to systematically validate 40 representative peptides through chemical synthesis and experiments. The validation pipeline included: in vitroantibacterial activity testing against Gram-positive and Gram-negative bacteria, hemolytic activity assessment, cytotoxicity testing on mammalian cells (CCK-8 assay), scanning electron microscopy observation of membrane disruption mechanisms, and ultimately evaluation of their in vivoantibacterial and wound-healing efficacy in a mouse infection model, forming a complete chain of evidence.
03 Results and Discussion
3.1 Outstanding Performance of the AMPLiT Tool
The final integrated AMPLiT model (Model G) demonstrated high sensitivity (90.13%) and specificity (99.69%) on the test set. Its AUPRC value (0.9486) was comparable to that of a more complex baseline model (AMPidentifier 1.0), but with significantly reduced training time and computational resource requirements, proving its efficiency.
3.2 High-Yield Discovery and In VitroValidation of Ancient AMPs
From seven ancient coprolite samples, after stringent decontamination and filtering pipelines, 160 non-redundant candidate AMPs were obtained. Filtering these further, 40 peptides were successfully synthesized for in vitrovalidation. Impressively, 36 of these peptides (90%) exhibited measurable growth inhibitory activity against at least one tested pathogen (Bacillus subtilis, Staphylococcus aureus, Escherichia coli, Pseudomonas aeruginosa) at concentrations of 100 µM or lower, demonstrating the high predictive accuracy of AMPLiT.
3.3 Genomic Context and Conservation of S. copri-derived AMPs
Analysis of the genomic loci encoding the active AMPs revealed that approximately 77% of the ORFs were fragments of larger genes encoding housekeeping proteins like topoisomerase-primases or the cell division protein FtsZ. This aligns with a microbial "resource-saving" evolutionary strategy of repurposing transmembrane segments of housekeeping genes to generate new AMPs. Although ancient DNA degradation complicated precise taxonomic assignment, many AMP ORFs were found to be conserved in modern S. coprigenomes from rural populations and showed limited overlap with sequences in known AMP databases, indicating their uniqueness.
3.4 Potent Activity and Safety Profile of Lead AMPs
Dose-response experiments showed that several S. copri-derived AMPs (e.g., AMP8, AMP13, AMP24, AMP30, AMP31, AMP38) exhibited potent inhibitory activity against various pathogens at concentrations between 5-100 µM. Safety assessments revealed these lead peptides had minimal hemolytic activity (<10%) even at 100 µM and showed no significant toxicity to intestinal epithelial cells (Caco-2). Scanning electron microscopy directly observed their ability to disrupt the integrity of pathogen cell membranes, confirming their mechanism of action.
3.5 In VivoEfficacy Validation
In a mouse wound infection model, topical application of selected AMPs (AMP8, AMP13, AMP30, AMP31, AMP38) significantly reduced the bacterial load of S. aureusand P. aeruginosaand accelerated wound healing. Histological analysis (H&E and Masson's staining) showed reduced immune cell infiltration and improved tissue repair in AMP-treated groups, with effects comparable to those treated with vancomycin or polymyxin B.
04 Conclusion and Future Perspectives
This study successfully demonstrates that ancient gut microbiomes, particularly the dominant S. copriwithin them, constitute a valuable resource for novel antimicrobial peptides. By developing the efficient, lightweight AI mining tool AMPLiT and conducting systematic experimental validation, the researchers have "resurrected" these "lost" antimicrobial molecules, which exhibit excellent antibacterial activity, favorable safety profiles, and in vivoefficacy comparable to clinical antibiotics. This work pioneers a new paradigm of "molecular de-extinction," seeking solutions to current medical challenges from the evolutionary history of the human microbiome. The decline of S. coprimight be a factor in modern human microbiome dysbiosis, and its derived AMPs could serve not only as novel antibiotic candidates but also as tools to restore microbiome ecological function. Future research could focus on: ① expanding the sampling of ancient specimens (different eras, cultures); ② confirming the natural expression of these AMPs in ancient environments using techniques like proteomics; ③ enhancing peptide stability via strategies like D-amino acid incorporation or cyclization; ④ systematically evaluating their efficacy against modern drug-resistant pathogens and their synergy with existing antibiotics; and ⑤ investigating their potential impact on human commensal flora and the immune system. The AMPLiT tool and the AMP sequences uncovered in this study will provide a crucial platform for researchers worldwide to continue exploring this valuable resource. In conclusion, this research strongly suggests that solutions to the antibiotic resistance crisis may lie hidden within the microbiomes of our ancestors.
Original Article：
Chen S, Yuan Y, Wang Y, Peng Y, Tun HM, Jiang Z, Miao Y, Lee S, Yin X, Shen X, DeLeon O, Chang EB, Chan FKL, Sun Y, Ng SC, Su Q. Identification of antimicrobial peptides from ancient gut microbiomes. Nat Commun. 2026 Jan 14.