Repository navigation
Node-fraction sweep (crc_232) and Celcomen baseline notebook - #59
Merged
Merged
Conversation
Node-perturbation fraction experiment: run_node_fraction.py (single-run inference worker, defaults now point at slide 232) and plot_per_ct.py (per-cell-type tables/figure, defaults to results_232), plus the shared loo_metrics.py scorer and the slide-232 results (5 seeds x 7 fractions x 5 cell types). Celcomen baseline notebook on crc_232 with its conda environment; it documents the O(cells^2) memory requirement that prevents training on this slide. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Rerun the slide-232 node-perturbation fraction sweep with cellina feat/perturb-fraction (PR PMBio/cellina#39) over fractions 0.1, 0.25, 0.5, 0.75, 1.0 (0.0 and 0.05 dropped), 5 cell types x 5 seeds, and regenerate the per-cell-type summary tables and figure. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Requires: cellina with
perturb_fractionsupport, PMBio/cellina#39.Summary
notebooks/additional/node_fraction/run_node_fraction.py: inference-only worker that applies the tutorial 4.2 node perturbation to a random fraction of neighbour cells using a pre-trained k=200 within-domain checkpoint. Defaults now target slide 232 (crc_232.h5ad,232_REF->232_CRC), data path resolved relative to the repo.notebooks/additional/node_fraction/plot_per_ct.py: aggregates<results>/<ct>/frac*_seed*.jsoninto the combined Pearson / L2 LaTeX tables, CSVs and the per-cell-type overlay figure. Defaults toresults_232.notebooks/additional/loo_metrics.py: shared scorer (imports metric implementations fromscripts/counterfactual_analysis.py).notebooks/additional/node_fraction/results_232/: the slide-232 results (5 seeds x 5 fractions x 5 cell types; fractions 0.1, 0.25, 0.5, 0.75, 1.0, with 0.0 and 0.05 omitted), rerun with cellinafeat/perturb-fractionplus the generated tables/figure. Runningplot_per_ct.pywith no arguments reproduces the committed.textables byte-for-byte.notebooks/additional/celcomen/celcomen_crc232.ipynb+environments/celcomen.yml: Celcomen run as released on the benchmark setting forcrc_232. Documents the dense cell-by-celltrace(msg @ gex.T)allocation (42.6 GiB at 107k cells) that makes training infeasible on this slide.Notes
run_node_fraction.pyneeds--ckpt-rootpointing at thek200_seed{seed}checkpoints trained by the graph-sensitivity sweep with the same cell type held out. That training script and therun_232.shdriver are not part of this PR.condition_on_intrinsic=Falsein their init params, and cellina v1.1.1 / 33 concert in ood bench #39 no longer accepts that kwarg. For this rerun the key was dropped from scratch copies of the checkpoints. That is equivalent, becauseFalseis the only code path left.🤖 Generated with Claude Code