Encoded expertise, not canned pipelines.
A skill pack is what a good computational biologist knows about a class of analysis: which methods fit which study designs, which assumptions break, which diagnostics to run before you believe a result. Inflexa ships more than twenty of them, and they are all open.
What a skill pack is, and what it is not
It is: guidance. A pack tells the agent how a competent analyst would approach this kind of problem, what to check first, which method is appropriate when the design is unbalanced, what a suspicious result looks like, and when to stop and say the data will not support the question.
It is not: a validated, certified, turnkey pipeline. Nobody has run a formal validation suite against these packs. They encode expertise, and expertise is fallible. Inflexa will show you every method it chose and why, precisely because you are still the one who has to judge whether it was the right call.
If a vendor tells you their analysis packs are validated, ask them what the validation consisted of. We would rather be the ones who told you the honest answer first.
Twenty-plus packs, and growing in the open
New packs land as pull requests, which means you can read them before they run, argue with them in an issue, and write your own.
Transcriptomics
- Bulk RNA-seq differential expression
- Single-cell RNA-seq
- Pathway and gene-set enrichment
- Transcription-factor activity inference
- Co-expression networks
Other modalities
- Proteomics
- Cheminformatics
- Imaging
- Multi-omics integration
- Cross-study meta-analysis
Inference and rigour
- Causal inference and confounding adjustment
- Survival analysis
- Batch-effect diagnosis
- Sensitivity and robustness testing
- Multi-method consensus
Translational
- Target prioritisation and tractability
- Toxicogenomics and safety
- Pharmacovigilance signal detection
- Drug repurposing
- Literature evidence chains
Your lab's method is a skill pack too.
Every group has a way of doing things: a normalisation step you insist on, a diagnostic you always run, a threshold you learned the hard way. That knowledge usually lives in one person's head and leaves when they do.
A skill pack is a place to put it. Write it once, and every analysis your group runs from then on knows what you know. Keep it private in your own fork, or open a pull request and let the rest of the field have it, that choice is yours, and it stays yours, because the licence is Apache 2.0.
Read a pack before you trust it.
They are plain text in a public repository. That is the whole point.