Supplementary Files SupplementaryFile1.csv.gz – Metadata for field collected samples with columns: “sample_id” – individual sample names. “health” – “H” healthy and “D” diseased fragments. “year” – year fragment collected. “season” – season fragment collected (“S” July and “W” January) “site” – location fragment collected from “lib.size” – total number of sequenced reads “norm.factors” – factor used to normalize read counts of ASVs SupplementaryFile2.csv.gz – Metadata for tank collected samples with columns: “sample_id” – individual sample names. “geno” – fragment genotype “fragment_id” – fragment identification tracked through repeated sampling “tank_id” – tank identification “time_treat” – concatenated metric for sampling time, exposure, and disease outcome separated by “_” Time – 0, 2, 8 Exposure – “D” Diseased, “N” Healthy Disease Outcome - “D” Diseased, “H” Healthy “lib.size” – total number of sequenced reads “norm.factors” – factor used to normalize read counts of ASVs SupplementaryFile3.fasta – FASTA file including complete 16s sequences named with ASV identifier and taxonomy. SupplementaryFile4.csv.gz – Matrix of the number of reads of each ASV sequenced in each sample. Combined both field and tank samples. SupplementaryFile5.csv.gz – Matrix of the log2 CPM of each ASV sequenced in each sample. Combined both field and tank samples. SupplementaryFile6.csv.gz – Complete results for each ASV association. “top_classification” – lowest taxonomic classification with more than 80% confidence. “taxonomy” – Full taxonomy including confidence in each taxonomic level. “passedFilter” – indicates taxa filtered from analysis due to rarity and/or lack of observations across sample times. “rank_*” – machine learning model rankings, median ranking, and model estimated ranking along with standard error, confidence interval, and FDR adjusted p-value used to identify important ASVs. “ml_retained” – Indicates if the ASV was of above average importance to ML models. NA values indicate ASVs which were filtered prior to ML modelling. “fieldModel_*” – ANOVA table results for each ASV testing the effects of health, year, season and all possible interactions indicating: Sums of squares, mean squares, numerator and denominator degrees of freedom, F statistic, p-value, and FDR corrected p-value. NA values are filled for ASVs filtered prior to differential abundance analysis. “diffAbundance_healthAssociation” – Marks the health association of ASVs from differential abundance analysis of field samples: “H” health, “D” diseased, “N” none, NA – filtered prior to differential abundance analysis. “fieldLogFC_*” – Post-hoc contrasts for ML retained ASVs testing the significance of the log2 fold-change between disease and healthy fragments within each sampling time (year: 2016, 2017 & season: “S” July, “W” January) showing: Mean estimate, standard error, degrees of freedom, lower and upper 95% confidence interval, t-statistic, p-value, FDR adjusted p-value. NA values are filled for ASVs which were not marked as important by ML models. “field_consistent” – Indicates if the ASV was consistently healthy or disease associated across sampling times. NA values are filled for ASVs which were not marked as important by ML models. “tankModel_*” – ANOVA table results for each ASV testing the effect of the combination of time, disease exposure, and disease outcome, indicating: Sums of squares, mean squares, numerator and denominator degrees of freedom, F statistic, p-value, and FDR corrected p-value. NA values are filled for ASVs filtered prior to tank experimental analysis. “tankLogFC_*” – Post-hoc contrasts for ASVs tested in tank exposure experiments. Contrasts include: Post-exposure diseased vs healthy outcome regardless of exposure (DvH) Post-exposure diseased vs healthy exposure regardless of outcome (DvN) Post-exposure disease exposed corals with disease symptoms compared to disease exposed but still healthy corals (DDvDH) Post-exposure disease exposed corals with disease symptoms compared to healthy exposed and still healthy corals (DDvNH) Post-exposure disease exposed corals which stay healthy compared to healthy exposed and still healthy corals (DHvNH) Pre-exposure compared to Post-exposure in corals with the disease regardless of exposure (PostvPreD) Pre-exposure compared to Post-exposure in corals without the disease regardless of exposure (PostvPreH) Mean estimate, standard error, degrees of freedom, lower and upper 95% confidence interval, t-statistic, p-value, FDR adjusted p-value. NA values are filled for ASVs which were not consistently associated with healthy or diseased corals in the field experiment. “pathogen_classification” – Indicates the predicted microbial classification based on the tank results. Pathogen, Opportunist, Commensal NA values are filled for ASVs which were not consistently associated with healthy or diseased corals in the field experiment.