Over the past decade, forensic microbiology has evolved from a nascent concept into a powerful tool for connecting crime scenes, suspects, and victims^[1,2]. Enabled by high-throughput sequencing, researchers can now delineate the microbial “fingerprints” on physical evidence or individuals with unprecedented resolution. By comparing the taxonomic composition and community structure between samples, significant strides have been made in human identification, geographic sourcing, and post-mortem interval estimation^[2-8]. The central tenet of this research has been to treat microbial communities primarily as stable, static taxonomic labels. This paradigm presumes that a microbial census, captured at a specific moment, faithfully reflects its origin and thus serves as the core basis for associative evidence.

However, this foundational view of microbes as static taxonomic identifiers fails to fully harness their informational potential. This conventional approach has inherent limitations in reconstructing the specifics of an event. First, while it acknowledges the dynamic responsiveness of organisms to environmental change^[9-11], its focus on community-level shifts often provides limited insight into the causal pressures driving these changes. For instance, analyses of taxonomic succession are valuable for estimating the post-mortem interval, but they may not distinguish between different types of stressors that could lead to similar community structures. Consequently, a static species list provides an incomplete record of critical events, such as scene sanitization or body disposal, which impose intense selective pressures sufficient to reshape the microbial genomic landscape within hours to days^[12-14]. Second, taxonomic-level comparisons can be confounded by ecological convergence, leading to ambiguity and diminishing the specificity and strength of the evidence^[15,16]. Therefore, while this approach is effective at addressing “who was here?”, it is often less equipped to resolve the more granular forensic question: “what happened here?”.

Here, we propose a paradigm shift to transcend these limitations: recasting microbes from passive “taxonomic labels” to active “dynamic environmental sensors” and “event recorders.” The core idea of this paradigm is that the microbial genome is not immutable but is a molecular ledger that responds to environmental stress in real time, leaving heritable alterations^[17]. When confronted with specific stimuli - such as the chemical assault of a disinfectant, the selective pressure of antibiotics, or contamination with heavy metals - microbes undergo adaptive evolution at the genomic level through mechanisms such as horizontal gene transfer (HGT), phage activity, or the rapid accumulation of point mutations^[18-24]. These genomic alterations, which we term “genomic scars”, are driven by environmental pressures and fixed by natural selection. In principle, these scars can document the specific events a microbial community has endured, offering a dimension of information far richer and more profound than a simple species roster.

In this Review, we will systematically elucidate this emerging field. We first delve into the key biological mechanisms that forge these “genomic scars”, including the roles of mobile genetic elements (MGEs), phage dynamics, and the accumulation of micro-scale genomic variations. Building on this foundation, we will construct a framework of potential forensic applications, demonstrating how decoding these scars can be used to reconstruct criminal actions, infer an individual’s lifestyle and provenance, and forge novel evidentiary links. Furthermore, we will highlight the key analytical technologies required to achieve these goals and, finally, discuss the challenges and future directions for the field. We contend that this paradigm shift - from taxonomic labels to event recorders - will open a new dimension for forensic science, marking a pivotal step for microbial evidence to evolve from being purely associative to becoming truly behavioral.

The concept of microbes as dynamic event recorders is scientifically grounded in a suite of rapid genomic adaptation mechanisms that microorganisms have evolved to survive drastic environmental shifts^[25-27]. These mechanisms enable microbial populations to transduce external physical or chemical stimuli into stable, heritable genomic alterations - what we define as “genomic scars” - on timescales ranging from hours to a few generations. However, the persistence of these scars is not indefinite, as their maintenance can impose a “fitness cost” on the organism once the stressor is removed. Therefore, understanding this temporal dynamic, or the “half-life” of a specific scar, is essential for accurately reconstructing the timeline of a forensic event and inferring how recently it occurred. These scars are not merely a testament to microbial survival strategies; they provide a layer of high-resolution information crucial for forensic reconstruction.

A paradigm shift that views microbial genomes as dynamic information carriers has opened up broad prospects for forensic applications, ranging from behavioral reconstruction to physical evidence association. These deeply embedded “genomic scars”, serving as direct responses to specific environmental stresses and events, can provide higher-dimensional and more targeted evidence compared to traditional species composition analysis. Building upon the robust foundations of existing research in forensic genetics, microbial ecology, evolutionary biology, and clinical microbiology, this chapter systematically elaborates on how to translate these “genomic scars” into actionable forensic practices, constructing three core application scenarios.

The paradigm shift of redefining microorganisms from mere “taxonomic labels” to “event recorders” imposes exceptionally high demands on analytical technologies. Traditional approaches such as 16S rRNA gene sequencing or shallow metagenomics are no longer sufficient. It is imperative to rely on a progressive technological framework comprising deep metagenomics, microbial single-cell genomics, and artificial intelligence (AI)-driven integrative analysis to transform microscopic “genomic scars” into forensic evidence characterized by high resolution, specificity, and interpretability.

Ultra-deep metagenomics overcomes the technical limitations of traditional metagenomics in detecting low-abundance species (relative abundance < 0.1%), resolving genomes at the strain level, and reconstructing complex microbial community structures by implementing ultra-deep sequencing (typically exceeding 50 Gb per sample) and integrating hybrid assembly strategies that combine long-read and short-read sequencing technologies^[64,65]. It not only enables the assembly of a large number of high-quality, nearly complete metagenome-assembled genomes (MAGs), but also systematically uncovers extrachromosomal MGEs (such as plasmids and bacteriophages), thereby providing a foundational framework for understanding the species composition, functional potential, and ecological interactions of microbial communities. Building upon this foundation, microbial single-cell genomics elevates the resolution to the level of individual cells. It not only reveals intercellular heterogeneity in chromosomal genomes but also directly links specific MGEs to their host cells^[66]. For example, it can precisely identify which specific strain or single cell harbors an antibiotic resistance plasmid, thereby accurately defining the carrier of functional genes. This is crucial for tracing the transmission pathways of functional genes within complex communities and identifying rare cells with key phenotypes (such as antibiotic resistance or unique metabolic capabilities). It provides direct evidence for enabling a shift from answering “who is there?” to “who is doing what?” However, the application of single-cell genomics to forensic samples, which are often characterized by low biomass and DNA degradation, poses significant challenges for isolating intact single cells. To overcome this limitation, proximity-ligation-based methods, such as meta-Hi-C (metagenomic chromosome conformation capture), offer a powerful complementary or alternative approach. This technology captures the physical proximity of DNA molecules within a cell, enabling the direct association of plasmids and phages with their host chromosomes even in complex samples where single-cell isolation is not feasible^[67,68]. Integrating such techniques would significantly enhance the reliability of linking event-specific genomic scars to the responsible microbial actors.

Ultra-deep metagenomics and single-cell genomics generate vast, multidimensional, and highly complex datasets. These data encompass not only multi-layered information such as species, genes, and functional pathways but also structural variations in chromosomes and MGEs, single-cell heterogeneity, and spatiotemporal dynamics. Traditional bioinformatics approaches face significant challenges in processing such high-dimensional, nonlinear, and noise-laden association networks, making it difficult to distinguish biologically meaningful and forensically valuable causal signals from complex correlations^[69]. AI technologies, particularly machine learning and deep learning, provide revolutionary tools for addressing the complexity of microbiome data. By constructing unified analytical frameworks that integrate multi-omics data, AI can automatically identify key microbial biomarkers, gene modules, or combinations of MGEs most relevant to specific forensic phenotypes - such as geographic origin, individual characteristics, and more - from vast microbial genomic datasets, thereby achieving dimensionality reduction and feature extraction^[55,70-72]. Furthermore, AI constructs potential interaction and transmission networks among microorganisms, between microorganisms and host genes, and of MGEs within communities. This approach moves beyond simple species abundance correlations to reveal the underlying ecological and evolutionary dynamics driving changes in microbial community structure. Ultimately, by applying advanced analytical frameworks such as causal discovery algorithms and structural equation modeling, AI provides critical support for reconstructing the temporal sequence and logical chain of events in forensic science^[73,74].

The paradigm of “microbial genomic scars” offers a potential pathway for forensic science to transition from identity association to behavioral reconstruction. Furthermore, a holistic approach that integrates the analysis of genomic scars with traditional community-level taxonomic shifts is crucial for building a robust interpretive framework. Genomic adaptations (the “cause”) and subsequent changes in community composition (the “effect”) are two sides of the same coin in a microbial ecosystem’s response to a stressor. For instance, identifying a specific antibiotic resistance plasmid (a genomic scar) within a bacterial strain, and concurrently observing the dramatic rise in that strain's abundance, provides a powerful, two-layered line of evidence. This integrated analysis not only strengthens the causal inference but also creates a more resilient and comprehensive narrative of the forensic event, moving beyond what either data type could reveal in isolation. However, its translation from a theoretical concept into reliable and admissible evidence in court faces three core challenges.

First is the fundamental challenge of signal discrimination. A critical hurdle is distinguishing genuine event-driven genomic signatures from the vast background noise generated by daily living and random genetic drift. This challenge begins at the crime scene itself; standardized pre-analytical protocols - from sampling methods and swab types to DNA preservation buffers - are crucial to ensure that detected genomic signals are true reflections of an event, not artifacts of evidence collection. Furthermore, the intensive sample preparation steps, such as whole-genome amplification and library construction, can introduce their own procedural artifacts, including polymerase chain reaction (PCR)-induced mutations or sequencing errors, which could be mistaken for genuine genomic scars. This risk is magnified in low-biomass or degraded forensic samples. Therefore, ensuring signal integrity requires not only standardized collection but also the implementation of ultra-high-fidelity methods and rigorous controls throughout the analytical process. Subsequently, the “genomic scars” of a forensic event must be robustly distinguishable from the vast background “noise” generated by both daily living and random genetic drift. For instance, individual-specific factors, such as personal hygiene habits (from fastidious to infrequent), diet, or medication use (e.g., a proton pump inhibitor), can introduce significant confounding effects. The strong signal induced by a sterilizer represents one end of a spectrum, while the subtle impact of a brief visit to a new location may be far more difficult to detect. Consequently, a unified statistical model applicable to complex communities is currently lacking to precisely distinguish genotypes driven by specific forensically relevant stimuli from this complex background of variation. This requires capabilities beyond mere variant detection, extending to the analysis of their allele frequency dynamics, genomic distribution patterns, and whether they exhibit signatures of a selective sweep.

Second is the challenge of establishing causality. Current research is largely limited to phenomenological descriptions, whereas the evidentiary value of this approach hinges on the critical leap from correlation to causation. There is an urgent need to construct a rigorously validated causal knowledge base of “stimulus-specific genomic scars” through controlled experiments. This would systematically elucidate the reproducible and predictable adaptive genomic changes induced by different stimuli.

Finally, a significant void exists in standardized evidentiary quantification. Unlike the mature probabilistic frameworks used in traditional DNA fingerprinting, evidence from microbial scars lacks a quantitative assessment system. Consequently, it cannot answer key questions regarding its evidential strength, the probability of its random occurrence, its degree of specificity, or the confidence level of temporal inferences.

To address these challenges, a concerted, multi-pronged effort is required in three key directions. First, developing novel algorithms that integrate evolutionary theory with machine learning. This involves combining population genetics models with deep learning to model mutation and selection processes at the strain level, thereby isolating genuine signatures of directional selection from background noise. Second, launching a large-scale, standardized “Microbial Forensic Stress-omics” initiative. Through international collaboration, this program would systematically chart a comprehensive atlas of “forensic stimulus-genomic response” profiles, establishing an open-access causal association database to provide a robust foundation for interpretation. Third, establishing a likelihood ratio evaluation framework and validation system for microbial evidence. Drawing on best practices from forensic genetics, this would involve building reference databases, developing statistical models to quantify evidential strength, and performing large-scale empirical validation through blind trials and retrospective case studies to establish scientific standards for admissibility in court.

Furthermore, it is imperative to contextualize microbial evidence within the broader ecosystem of forensic investigation. Forensic scenes are inherently complex, and no single piece of evidence, microbial or otherwise, can provide a complete picture. Therefore, the insights derived from genomic scars should not be interpreted in isolation. Instead, they must be integrated with traditional physical evidence (e.g., fingerprints, DNA profiles), chemical analyses, and circumstantial evidence. This multi-modal approach, where microbial findings corroborate or challenge other evidentiary lines, is essential for building a robust and legally defensible case. The ultimate goal is to weave microbial narratives into the larger tapestry of the investigation, using them as a powerful auxiliary tool to enhance, rather than replace, established forensic methodologies.

In summary, the ultimate success of this paradigm hinges on a rigorous and systematic resolution of the serial bottlenecks from signal discrimination and causal establishment to evidentiary quantification. Through interdisciplinary integration and methodological innovation, this field holds the promise of providing judicial practice with “microscopic time capsules” and “molecular narrators” capable of reconstructing the sequence of events.