Analysis of unmapped RNA-seq data from cancer spatial transcriptome toward characterizing cancer microbiome

Spatial characterization of microbial-like signals in tumor tissues remains challenging, particularly in direct Visium data, where microbial reads are sparse and may not be fully retained in standard count matrices. Here, we present an extended unmapped-read analysis as a proof-of-concept workflow for summarizing microbial-like 16S rRNA signals in four direct Visium specimens from colorectal cancer (CRC), oral squamous cell carcinoma (OSCC), and head and neck squamous cell carcinoma (HNSC). The workflow uses a custom reference containing four selected 16S rRNA sequences and computes a per-spot mismatch ratio to quantify sequence-level dissimilarity relative to each reference. Compared with PathSeq, the workflow yielded different spatial signal patterns and mismatch summaries across the analyzed specimens. Among the four tested references, the CRC specimen showed lower mismatch ratios relative to the E. coli reference than the other analyzed specimens, an observation compatible with the intestinal context but not definitive evidence of species-level presence or evolutionary proximity. Given the small sample set, restricted reference panel, and lack of dedicated negative controls, these findings should be interpreted as hypothesis-generating. This study provides a complementary proof-of-concept framework for exploring microbial-like signals in direct Visium data.