Bioscience research router

Bioscience Research Router

Route the task to the smallest sufficient scientific workflow, then enforce evidence, reproducibility, privacy, and human-review gates.

Route the work

Need	Primary skill	Combine with
Choose or rescue a research problem	scientific-problem-selection	scientific-literature-review, experimental-design
Review biomedical evidence	scientific-literature-review	shared deep-research when available
Plan controls, randomization, blocks, batches, or DOE	experimental-design	statistics/power skill from the shared master when needed
Design or validate PCR/qPCR primers or probes	molecular-genetics-assay-planning	scientific-literature-review for assay evidence
Set up or repair an RStudio project, renv, or a clean reproducible run	rstudio-reproducible-analysis	quarto-authoring for the report
Clean, join, reshape, or validate lab/omics tables in R	r-bioscience-data-wrangling	rstudio-reproducible-analysis
Choose, fit, diagnose, and report an R statistical model or contrast	r-biostatistics-workflow	experimental-design, r-bioscience-data-wrangling
Bulk RNA-seq differential expression from a count matrix	bio-differential-expression-deseq2-basics	r-biostatistics-workflow; nextflow-development upstream
Publication-quality figures in R with ggplot2	bio-data-visualization-ggplot2-fundamentals	r-biostatistics-workflow first
Reproducible scientific report (.qmd) with code, tables, and citations	quarto-authoring	rstudio-reproducible-analysis
Thesis defense, seminar, lab meeting, grant, or conference presentation	academic-presentations	bio-data-visualization-ggplot2-fundamentals, scientific-literature-review
Synthesize interviews, notes, findings, or mixed evidence	research-synthesis	scientific-literature-review for source-backed literature
Organize durable paper, method, protocol, or project notes	knowledge-base	quarto-authoring for reproducible reports
QC .h5ad or 10x .h5 single-cell RNA data	single-cell-rna-qc	scvi-tools after QC
Integrate or model single-cell/multiome data	scvi-tools	single-cell-rna-qc first
Run reproducible RNA-seq, WGS/WES, or ATAC-seq pipelines	nextflow-development	experimental-design, literature review
Check U.S. IRB/IBC/GLP/GMP/FDA context	this skill	read references/regulatory-routing.md

Do not invoke every skill. Use one primary skill and only the supporting skills that close an identified gap.

Scientific execution contract

1. State the scientific question, biological unit, sample type, organism, reference build, data modality, and intended decision.
2. Separate raw data, derived data, code, parameters, and interpretation. Never overwrite raw data; work on a copy and preserve identifiers or checksums.
3. Record controls, biological versus technical replicates, randomization, blocking, batch structure, exclusion rules, and planned statistics before interpreting results.
4. Verify actual tool and package versions before using examples. Pin an exact tested version for reproducible runs; do not silently upgrade a working pipeline.
5. Run a small validation or test profile before full data. Preserve logs, commands, seeds, environment information, and failure messages.
6. Report evidence, assumptions, uncertainty, alternatives, and validation status separately. Label hypotheses and preprints explicitly.
7. Require a qualified human to approve biological interpretation, wet-lab execution, regulated decisions, clinical relevance, and publication claims.

Read references/evidence-and-reproducibility.md for the evidence hierarchy and minimum audit trail.

Privacy and external systems

• Treat unpublished sequences, patient-linked data, sample identifiers, credentials, and institutional records as sensitive.
• Prefer local processing. Do not upload sensitive or controlled data to a third-party service without explicit authorization and the institution's approved data-use path.
• Remove direct identifiers and unnecessary metadata before sharing. De-identification does not itself prove HIPAA or institutional compliance.
• Use public accession downloads only after confirming the accession and intended dataset. Record the source URL, accession, retrieval date, and checksum when available.

Safety and compliance boundary

• Provide scientific and regulatory source navigation, not medical, legal, biosafety, IRB, IBC, GLP, GMP, or quality-unit approval.
• Do not decide whether an activity is exempt, compliant, safe, diagnostic, or ready for clinical/manufacturing use.
• For human participants or identifiable biospecimens, stop at a draft/research plan and route the decision to the institution's IRB/privacy office.
• For recombinant or synthetic nucleic acids, engineered organisms, viral vectors, pathogens, toxins, or uncertain containment, route the protocol to the institutional IBC/EHS/biosafety officer before execution.
• For animal work, route design and procedures to the institution's IACUC/veterinary oversight.
• For regulated nonclinical or manufacturing work, distinguish research guidance from controlled SOPs and route final decisions to QA/regulatory professionals.

Read references/regulatory-routing.md whenever any regulated trigger appears.

Deliverable

Return a compact research packet:

• question and scope;
• selected workflow and why;
• inputs and provenance;
• analysis or experiment plan;
• controls and validation gates;
• evidence table with citations;
• assumptions, uncertainties, and alternatives;
• human approvals or institutional reviews required;
• reproducibility record and next safe action.