Add HDF5-backed CytoTalk cached runner script

[pooriass]

Summary

• add an R script that builds a file-backed mutual information cache for CytoTalk runs
• implement a streamed ARACNE DPI pruning routine to avoid loading full MI matrices into RAM
• provide a cached runner helper and example invocation for a cell-type pair

Testing

• not run (not applicable)

---

https://chatgpt.com/codex/tasks/task_b_68f2390e87dc8320a8e5e3af4d5d8d3b

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting

• @codex fix this CI failure
• @codex address that feedback

[chatgpt-codex-connector]

Iterate neighbor pairs without producing invalid indices

The inner DPI loop iterates for (b in (a + 1L):length(idx_nz)). In R the x:y operator returns a decreasing sequence when x > y, so when a reaches the final neighbor the loop yields indices greater than length(idx_nz) (e.g. 3:2 → 3, 2). Those out‑of‑range b values make kept[b] evaluate to NA, which triggers if (!kept[b]) to error with “missing value where TRUE/FALSE needed” as soon as a node has two or more neighbors. As a result, the default streaming ARACNE path fails on practically all nontrivial datasets. The loop needs to guard against a == length(idx_nz) (e.g. iterate only when a < length(idx_nz) or use seq.int(a + 1L, length(idx_nz)) with a check).

Useful? React with 👍 / 👎.