Scientific tool catalog
Use this catalog to find the Open-Science entry point and runtime requirements for a scientific method. It is a documentation index, not an application page that installs every listed program.
Software families and where they run
| Software family | Open-Science entry | Setup to verify | Useful first check |
|---|---|---|---|
| Python standard library and plotting | Session Notebook | Selected Python interpreter; plotting dependencies | Read the real QC CSV and produce a small plot |
| R / base R | Session Notebook; Settings → Runtimes → R | Enabled app-managed or detected interpreter | Print R version and reproduce sample summaries |
| Structure-prediction frameworks | Relevant bundled Skill, usually Compute | Compatible GPU, packages, weights, input format and external MSA access if used | One small valid sequence/complex and its output confidence |
| MPNN sequence-design programs | ProteinMPNN / LigandMPNN / SolubleMPNN Skills | Repository/checkpoints and Python dependencies; small CPU jobs are supported by these instructions | One backbone with explicit fixed/design positions |
| Single-cell frameworks | scGPT / scvi-tools Skills | AnnData, cell/gene labels, packages and suitable compute | Validate input dimensions and required layers before training |
| Molecular structure rendering | Molecule Connector / molecule viewer | Built-in offline OpenChemLib route | Save and reopen aspirin.mol |
| File renderers | File preview | Supported extension and preview-size limits | Open the actual downloaded file |
| Remote batch software | Compute Host and remote-compute Skills | Host access, scheduler and named environment | Host probe followed by a bounded job |
See Skill directory for the complete 23-Skill method table. This page explains software readiness; it does not duplicate every Skill's procedure.
Inspect the selected environment
Use Settings → Runtimes to inspect available Python/R interpreters. Select the runtime actually bound to the session, then inspect packages there. An executable installed elsewhere on the computer is not automatically the active Notebook interpreter.
Use Runtimes to prepare an interpreter and Notebook to verify a calculation. If additional packages are needed, check installation and import in that environment. For a download failure, follow Network using the affected hostname and error.
Before following a model's install command
Read the installed Skill's exact requirements and the selected environment's supported setup route. Check package-name collisions: fair-esm and the Biohub esm implementation are distinct even though they share the esm namespace. A downloaded model's code and weights can also have different versions and access conditions.
For a local Notebook use the supported package-management flow in Scientific tools. For a remote host use Remote compute. A renderer displaying a PDB proves that it can view a structure; it does not prove AlphaFold or another prediction program is installed.
Implementation reference: manifest.json, SKILL.md, notebook-runtime.ts.
Use Background tasks to track supported long-running work and inspect delivered results. Scientific packages still need to be available in the selected runtime.