Update reporting#7
Merged
Merged
Conversation
Make the generation axis a fair comparison rather than a Stack-only run: - build_additive_deltas: each drug's mean real L1000 delta, applied to every organoid (organoid-independent) -- the generation analogue of the drug-mean floor. - build_learned_deltas: PCA/NMF organoid-specific delta predictors (a linear baseline->delta-residual map fit on real L1000, applied to each organoid's baseline), between the additive floor and Stack. - build_l1000_gdsc_pairs now also returns the per-cell-line DMSO baseline the learned predictors need. - regret_norm_at_k: shared clinical regret@k metric. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
score_viability_adapters runs the full {delta source} x {readout adapter} grid
against Soragni viability with global/interaction/within-drug rho, regret@k, and
the within-drug permutation null. Sources: additive (always), PCA/NMF learned
predictors (when the Soragni baseline is present), and Stack-generated (when
--generated-dir is given, now optional). validate_bridge_readout updated for the
new build_l1000_gdsc_pairs return.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Drop the encoder/probe (head-invariance) reporting -- it tested Stack only as a frozen embedding extractor on off-distribution pseudo-cell input, not its native mode. Reframe around the prompt->query generation axis: single-axis flowchart (L1000 context + Soragni query -> Stack generation / additive / PCA-NMF -> readout adapters -> metrics -> controls), data transformations, models, controls, regret, and the Path-B result with numbers explicitly pending an Alpine run. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Switch every Soragni build_sample_design call (and the prep_stack_input / benchmark_soragni --rna-source defaults) from organoid to tumor RNA, and update the README to match. The drug-screen response (the y label) is unchanged -- screens are on organoids -- so the prediction becomes the clinically realistic cross-substrate task: predict the organoid drug response from the patient tumor transcriptome. Tumor RNA covers all 17 matched patients, so the cohort is unchanged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GDSC2/DepMap RNA-seq is no longer a model input (only GDSC2 AUC, as readout training labels), so drop the GDSC2 cohort-composition panel and the GDSC2<->Soragni shared-panel figure (transfer-axis leftovers). plot_data now emits: Soragni-only cohort composition, response distributions (Soragni viability = target; GDSC2 AUC relabeled as readout labels), the Soragni organoid x drug viability heatmap, and a new L1000 drug-coverage figure (which Soragni drugs have a real L1000 perturbation). README figure block updated to match. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GDSC2 AUC no longer earns a figure -- it's only readout training labels, not a raw input worth plotting. The viability figure is now Soragni-only (the prediction target). Replace the L1000 coverage plot with a printed text list (and a README list): 18/26 Soragni drugs have an L1000 perturbation, so only those are scorable on the generation axis. Co-Authored-By: Claude Opus 4.8 (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
1 participant
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.
Updated the reporting to report the use of the tumor RNA-seq as the evaluator input. Also, Added scripts to score the adapters.