Machine learning (ML) has shown great promise in genetics and genomics where large and complex datasets have the potential to provide insight into many aspects of disease risk, pathogenesis of genetic disorders, and prediction of health and wellbeing. However, with this possibility there is a responsibility to exercise caution against biases and inflation of results that can have harmful unintended impacts. Therefore, researchers must understand the metrics used to evaluate ML models which can influence the critical interpretation of results. In this review we provide an overview of ML metrics for clustering, classification, and regression and highlight the advantages and disadvantages of each. We also detail common pitfalls that occur during model evaluation. Finally, we provide examples of how researchers can assess and utilise the results of ML models, specifically from a genomics perspective.

Characterizing the dynamics of microbial community succession in the infant gut microbiome is crucial for understanding child health and development, but no normative model currently exists. Here, we estimate child age using gut microbial taxonomic relative abundances from metagenomes, with high temporal resolution (±3 months) for the first 1.5 years of life. Using 3154 samples from 1827 infants across 12 countries, we trained a random forest model, achieving a root mean square error of 2.56 months. We identified key taxonomic predictors of age, including declines in Bifidobacterium spp. and increases in Faecalibacterium prausnitzii and Lachnospiraceae. Microbial succession patterns are conserved across infants from diverse human populations, suggesting universal developmental trajectories. Functional analysis confirmed trends in key microbial genes involved in feeding transitions and dietary exposures. This model provides a normative benchmark of “microbiome age” for assessing early gut maturation that may be used alongside other measures of child development.

Chitinases are important enzymes that contribute to the generation of carbon and nitrogen from chitin, a long chain polymer of N-acetylglucosamine that is abundant in insects, fungi, invertebrates and fish. Although mammals do not produce chitin, chitinases have been identified in bacteria that are key virulence factors in severe respiratory, gastrointestinal and urinary diseases. However, it is unclear how these enzymes are able to carry out this dual function. Legionella pneumophila is the causative agent of Legionnaires' disease, an often-fatal pneumonia and its chitinase ChiA is essential for the survival of L. pneumophila in the lung. Here we report the first atomic resolution insight into the pathogenic mechanism of a bacterial chitinase. We derive an experimental model of intact ChiA and show how its N-terminal region targets ChiA to the bacterial surface after its secretion. We provide the first evidence that L. pneumophila can bind mucins on its surface, but this is not dependent on ChiA. This demonstrates that additional peripheral mucin binding proteins are also expressed in L. pneumophila. We also show that the ChiA C-terminal chitinase domain has novel Zn²⁺-dependent peptidase activity against mammalian mucin-like proteins, namely MUC5AC and the C1-esterase inhibitor, and that ChiA promotes bacterial penetration of mucin gels. Our findings suggest that ChiA can facilitate passage of L. pneumophila through the alveolar mucosa, can modulate the host complement system and that ChiA may be a promising target for vaccine development. A broad range of organisms produce chitinase enzymes that digest chitin, the second most abundant carbohydrate on earth. Chitinases have also been identified that are important factors in major bacterial diseases but it is unclear how. Legionella pneumophila causes Legionnaires’ disease, a severe form of pneumonia, and its chitinase ChiA is essential for the survival of L. pneumophila during infection of the lung. Using structural biology and microbiology methods we have determined that ChiA can associate with the L. pneumophila surface and along with other outer membrane proteins can also bind mammalian mucins. We also identified a new and unique enzyme mechanism where L. pneumophila ChiA can hydrolyse the peptide bonds of mucin-like proteins. Mucins are major components of the mucous that lines the surface of the respiratory, digestive, and urogenital tracts and acts as a first line of defence during infection. This is the first mechanistic understanding of how a chitinase can promote disease through additional peptidase activity and suggests that L. pneumophila ChiA can modulate host immune responses and disperse host mucosa during infection.

The human gut microbiota is of increasing interest, with metagenomics a key tool for analyzing bacterial diversity and functionality in health and disease. Despite increasing efforts to expand microbial gene catalogs and an increasing number of metagenome-assembled genomes, there have been few pan-metagenomic association studies and in-depth functional analyses across different geographies and diseases. Here, we explored 6014 human gut metagenome samples across 19 countries and 23 diseases by performing compositional, functional cluster, and integrative analyses. Using interpreted machine learning classification models and statistical methods, we identifiedFusobacterium nucleatumandAnaerostipes hadruswith the highest frequencies, enriched and depleted, respectively, across different disease cohorts. Distinct functional distributions were observed in the gut microbiomes of both westernized and nonwesternized populations. These compositional and functional analyses are presented in the open-access Human Gut Microbiome Atlas, allowing for the exploration of the richness, disease, and regional signatures of the gut microbiota across different cohorts.

Objective: Targeting bacterial translocation in cirrhosis is limited to antibiotics with risk of antimicrobial resistance. This study explored the therapeutic potential of a non-absorbable, gut-restricted, engineered carbon bead adsorbent, Yaq-001 in models of cirrhosis and acute-on-chronic liver failure (ACLF) and, its safety and tolerability in a clinical trial in cirrhosis. Design: Performance of Yaq-001 was evaluated in vitro . Two-rat models of cirrhosis and ACLF, (4 weeks, bile duct ligation with or without lipopolysaccharide), receiving Yaq-001 for 2 weeks; and two-mouse models of cirrhosis (6-week and 12-week carbon tetrachloride (CCl4)) receiving Yaq-001 for 6 weeks were studied. Organ and immune function, gut permeability, transcriptomics, microbiome composition and metabolomics were analysed. The effect of faecal water on gut permeability from animal models was evaluated on intestinal organoids. A multicentre, double-blind, randomised, placebo-controlled clinical trial in 28 patients with cirrhosis, administered 4 gr/day Yaq-001 for 3 months was performed. Results: Yaq-001 exhibited rapid adsorption kinetics for endotoxin. In vivo , Yaq-001 reduced liver injury, progression of fibrosis, portal hypertension, renal dysfunction and mortality of ACLF animals significantly. Significant impact on severity of endotoxaemia, hyperammonaemia, liver cell death, systemic inflammation and organ transcriptomics with variable modulation of inflammation, cell death and senescence in the liver, kidneys, brain and colon was observed. Yaq-001 reduced gut permeability in the organoids and impacted positively on the microbiome composition and metabolism. Yaq-001 regulated as a device met its primary endpoint of safety and tolerability in the clinical trial. Conclusions: This study provides strong preclinical rationale and safety in patients with cirrhosis to allow clinical translation. Trial registration number: NCT03202498 .

Malnutrition affects over 30 million children annually and has profound immediate and enduring repercussions. Survivors often suffer lasting neurocognitive consequences that impact academic performance and socioeconomic outcomes. Mechanistic understanding of the emergence of these consequences is poorly understood. Using multi-system SHAP interpreted random forest models and network analysis, we show that Moderate Acute Malnutrition (MAM) associates with enrichment of faecal Rothia mucilaginosa, Streptococcus salivarius and depletion of Bacteroides fragilis in a cohort of one-year-old children in Dhaka, Bangladesh. These microbiome changes form interconnected pathways that involve reduced plasma odd-chain fatty acid levels, decreased gamma and beta electroencephalogram power in temporal and frontal brain regions, and reduced vocalization. These findings support the hypothesis that prolonged colonization by oral commensal species delay gut microbiome and brain development. While causal links require empirical validation, this study provides insights to improve interventions targeting MAM-associated neurodevelopmental deficits.