4.8 Typical Use Cases
4.8.1 Complete Metagenomic Analysis
User: “I uploaded paired-end FASTQ files, a host reference genome, and metadata.tsv. Discover a suitable BioLine Flow, check its inputs and parameters, and show me the plan. Submit only after I confirm. After upstream success, find a suitable downstream Tool for taxonomic profiling, group comparison, figures, and reporting.”
AI: “I will use the current BioLine Flow list and inspect responses to identify candidates, then show the Flow/Version ID, bindings, and defaults. I will create a Job only after your confirmation. Once upstream output is finalized, I will check an exact downstream Tool version and its resources, then pull it on demand and generate a task script.”
This covers input checks, Flow discovery, parameter confirmation, Job monitoring, output finalization, and recording the downstream Tool version and script.
4.8.2 Downstream-Only Analysis of an Existing Abundance Table
User: “I have a Bracken abundance table and metadata.tsv and do not need upstream processing. Discover an available downstream BioLine Tool, then perform statistics, differential analysis, and reporting.”
Check sample and group correspondence first, then verify the selected Tool summary/config, input format, and resource requirements. Pull that exact version on demand and analyze the existing data with a task-specific script.
4.8.3 Deep-Learning Inference
User: “Use a microbiome abundance table for phenotype prediction, or protein sequences for deep-learning inference.”
Discover a BioLine Tool compatible with the objective and input format. Check whether the exact version supports GPU, what resources it requires, and how its output can be interpreted. If no suitable version exists, do not infer capability from a historical image name; use the separate Tool authoring workflow.
4.8.4 Amplicon Analysis
User: “My inputs are 16S or ITS amplicon data. I want ASVs, taxonomy and functional prediction, followed by statistical analysis.”
Inspect the BioLine Flow and downstream Tool interfaces for input, method, port, and parameter support. Plan steps such as DADA2, QIIME2, taxonomy assignment, or functional prediction only when supported by the actual interface. If no suitable platform object exists, create and validate one on demand.