Machine-learning guided discovery of emergent antimicrobial activity from dynamic covalent assemblies（2b-RAD-M）

This dataset includes 24 biological samples: six healthy controls, six PBS-treated caries controls, six A5B5-treated samples and six chlorhexidine (CHX)-treated samples. Abundance.filtered.anno.csv contains all 24 samples in one annotated species-by-sample relative-abundance matrix. sample_metadata.csv provides the corresponding sample groups and original sample identifiers.</p><p dir="ltr"><br></p><p dir="ltr">Methods documented for the original samples: Library preparation and sequencing were conducted by OE Biotech (Qingdao, China) using a Type IIB restriction enzyme-based 2bRAD-M workflow, including BcgI digestion, adaptor ligation, PCR amplification, barcode indexing, and sequencing on the Illumina NovaSeq PE150 platform.

For downstream analysis, quality-controlled 2bRAD tags were mapped to a species-resolved 2bRAD marker database constructed from GTDB and Ensembl microbial genomes; candidate taxa were filtered using a Gscore threshold (>5) to reduce false positives, and species-level relative abundances were computed based on mapped reads and species-specific marker counts according to published 2bRAD-M protocols.</p>