1.3 End-to-End Experience: From Raw Data to Publishable Results

Using BioLine does not require the user to write code. The user uploads sequencing data and states the research objective in natural language—for example: “Compare salivary microbiome and metabolome differences between healthy controls and patients with multiple sclerosis, with particular attention to hypotaurine.” BioLine then organises the work according to the appropriate standard workflow.

The system does not immediately launch expensive computation. It first performs a pre-run check to verify file completeness, sample-to-group mapping, the proposed analysis plan, and the parameter list. Computation starts only after the user confirms the plan. Once submitted, the task is checked automatically at regular intervals. When upstream analysis finishes, BioLine verifies the completeness of the outputs before starting downstream diversity analysis, differential testing, multi-omics integration, figure generation, and report writing.

After completion, BioLine produces a reproducibility package containing tool versions, actual parameters, customised analytical scripts, container-environment definitions, and a one-command rerun script. Months or years later, the original researcher or another investigator can use the package to reproduce the analysis, including the generated figures.

BioLine currently supports five mature categories of analysis: full end-to-end metagenomics, upstream-only preprocessing, downstream-only reanalysis of abundance tables, 16S/ITS amplicon analysis, and deep-learning functional prediction such as disease classification, antimicrobial-resistance gene identification, and antimicrobial-peptide prediction.