Prediction of sea lice burden in Atlantic salmon using skin microbiome

This study evaluates the predictive capacity of the skin microbiota for sea lice burden in Salmo salar (Atlantic salmon) by leveraging machine learning and Hierarchical Feature Engineering (HFE). A total of 113 individuals were randomly selected, and skin mucus samples were collected via swabbing. The V4 region of the 16S rRNA gene was amplified and sequenced using the Illumina 250 paired-end protocol to identify amplicon sequence variants (ASVs) present within the bacterial community. Following an acclimation period, fish were experimentally exposed to C. rogercresseyi for 10 days. Parasitic load was quantified, and individuals were categorized based on parasite count using Random Forest (RF) algorithms, with and without HFE integration.

本研究借助机器学习与层级特征工程（Hierarchical Feature Engineering，HFE），评估了大西洋鲑（Salmo salar）皮肤微生物群对其海虱负荷的预测能力。研究共随机选取113尾个体，通过拭子采集皮肤黏液样本；对16S核糖体RNA基因的V4区进行扩增，并采用Illumina 250双端测序流程完成测序，以鉴定细菌群落内的扩增子序列变异体（amplicon sequence variants，ASVs）。经过驯化期后，将实验鱼暴露于罗杰西氏鱼虱（C. rogercresseyi）环境中10天。随后对寄生负荷进行定量，并基于寄生虫计数对个体进行分类，分别使用集成层级特征工程与未集成层级特征工程的随机森林（Random Forest，RF）算法开展分析。