Abstract
Microbial dynamics play an important role in larval health in aquaculture, but practical early warning tools remain limited. Here, we integrate flow cytometry (FCM) and 16S rRNA gene sequencing to identify microbial indicators that anticipate survival outcomes in barramundi (Lates calcarifer) hatcheries. From sequencing data, 54 amplicon sequence variants (ASVs) were found to differ consistently between high and low survival tanks. Machine-learning models trained on FCM phenotypic fingerprints (n=134) accurately inferred the abundances of selected ASVs and reproduced their survival-linked patterns across 20 independent larviculture batches (n = 1,206). Among them, four detrimental ASVs: Citrobacter (asv28, asv78), Rhodopseudomonas (asv6), and Arcobacter (asv47), emerged as predictors of poor survival, with discriminatory signals detectable as early as day 0 post-hatch (DPH 0). Classification models based on these markers at DPH 0 achieved moderate accuracy (AUC up to 0.74), demonstrating that early microbial community signatures can anticipate survival trajectories before mortality differences manifest at DPH 4-5. Importantly, direct FCM based classifiers also captured sufficient discriminatory information without sequencing input, highlighting a practical route for rapid, high-throughput monitoring. Together, these results position FCM-sequencing integration as a powerful approach to derive microbial survival markers and provide operational diagnostics to support proactive management in fish hatcheries.