GRAPH-DRIVEN WASTEWATER GENOMICS: ML IMPUTATION, EXPOSURE NETWORKS, AND CRISPR-VIRUS DYNAMICS - PROJECT SUMMARY MUNICIPAL WASTEWATER IS A UNIQUE AND UNDER-UTILIZED POPULATION-LEVEL RECORD OF PATHOGEN BURDEN, ANTIMICROBIAL- RESISTANCE (AMR) GENE FLOW, AND VIRUS-HOST CO-EVOLUTION. HOWEVER, SHORT-READ ASSEMBLIES AND FAILURE TO CONENECT WITH PUBLIC METADATA OBSCURE THESE SIGNALS AND DELAY OUTBREAK RESPONSE. THIS PROPOSED RESEARCH WILL BUILD A GRAPH-BASED COMPUTATIONAL FRAMEWORK THAT CONVERTS RAW WASTEWATER METAGENOMES INTO TIMELY, ACTIONABLE INDICA- TORS OF DISEASE. IN OUR FIRST AIM, WE WILL FOCUS ON GENOMIC SIGNAL RESCUE AND FUNCTIONAL ANNOTATION. WE WILL DEVELOP PATHOGEN- SPECIFIC COLORED DE BRUIJN SUBGRAPHS SEEDED BY KRAKEN AND METAPHLAN READS, THEN APPLY MASKED K-MER DIFFUSION TO PATCH READ GAPS AND RECOVER GENE-LEVEL BREADTH ACROSS VIRUSES, AIR-QUALITY-RESPONSIVE BACTERIA, AND EMERGING FUNGI (FOR EXAMPLE, CANDIDA AURIS). WE WILL NEXT USE THE RESCUED CONTIGS TOGETHER WITH DIRECT READ MAPPING TO (I) QUANTIFY CRISPR-SPACER TURNOVER, (II) CALL CORE-GENE VARIANTS, AND (III) FOLD-PREDICT NOVEL PROTEIN ALLELES, ANNOTATING ACTIVE SITES WITH FOLDSEEK. THIS WILL PRODUCE BREADTH AND DEPTH NORMALIZED COUNT TABLES, VARIANT LISTS, STRUCTURE ANNOTATIONS, AND A CONFIDENCE-BASED FALLBACK TO PLAIN MAPPING WHEN BREADTH GAIN IS LOW. THESE ENDPOINTS CLOSE THE GAP OF FRAGMENTED SHORT-READ COVERAGE THAT WASTEWATER METAGENOMICS ROUTINELY SUFFERS FROM WHILE PROVIDING THE FIRST GRAPH-AWARE, CROSS-KINGDOM CATALOGUE OF ADAPTIVE MUTATIONS IN WASTEWATER. IN OUR SECOND AIM, WE WILL FOCUS ON FORECASTING AND CAUSAL INFERENCE. USING EITHER OUTPUTS FROM FIRST AIM OUTPUTS OR PUBLICLY HARMONIZED COUNTS, WE WILL FIT ZERO-INFLATED NEGATIVE-BINOMIAL (ZINB) MIXED MODELS TO ISSUE OUTBREAK ALERTS FOR THE SAME PATHOGEN PANEL. WE WILL THEN APPLY LOCAL CAUSAL-GRAPH (DAG) DISCOVERY TO ENVIRONMENTAL COVARIATES DRAWN FROM NOAA CLIMATE DATA ONLINE (CDO), USGW NWIS WATER-DATA APIS, EPA AIR-QUALITY SYS- TEMS (AQS) API, AND NHANES. THESE PUBLICALLY AVAILABLE DATA SOURCES PROVIDE WEATHER, SEWER FLOW-RATE, AND AIR QUALITY DATA BETWEEN REAL-TIME TO HOURLY TIME POINTS. THESE WILL BE CONNECTED TO LAGGED LATENT ABUNDANCES TO REVEAL THE MINIMAL DRIVERS OF OBSERVED SURGES. FORECAST SKILL AND INFERRED EDGES WILL BE BENCHMARKED AGAINST MATCHED CLINICAL-INCIDENCE RECORDS, YIELDING AN INTERPRETABLE MAP OF TRANSMISSION ROUTES. THIS AIM CLOSES THE GAP LEFT BY ISOLATED, SINGLE-PATHOGEN DASHBOARDS BY UNIFYING VIRAL, BACTERIAL, AND FUNGAL TRENDS IN A STATISTICALLY RIGOROUS, CAUSALLY INFORMED EARLY-WARNING SYSTEM. ALL AIMS WILL BE EVALUATED ACROSS SEVERAL SUBSETS OF WEEKLY TIMEPOINT DATASETS: BIOTIA MUNICIPAL METAGENOMES, VERILY NWSS, AND SOUTH FLORIDA RADX-RAD SCHOOL-SITE LIBRARIES. THESE EACH OFFER DISTINCT SAMPLING PROTOCOLS AND DIFFERING METADATA. BY INTEGRATING GRAPHICAL METHODS TO PATCH CONTIG LENGTH WITH FUNCTIONAL ANNOTATION AND EXPLOR- ING DYNAMIC ENVIRONMENTAL MODELLING WITH THESE DATASETS, WE WILL DEVELOP A NOVEL WARNING SYSTEM OF EMERGENT INFECTION BURDEN AND PREDICTION, STRENGTHENING OUTBREAK PREPAREDNESS AND GUIDING TARGETED INTERVENTIONS.