Exact matrix-free streaming enables global phosphosite co-variation mapping without quadratic result materialization

Abstract

Reproducibility package for the Scientific Reports revision of the manuscript “Exact matrix-free streaming enables global phosphosite co-variation mapping without quadratic result materialization.” This clean-room package reproduces the principal computational and empirical results of the manuscript: an exact, feature-order-invariant global Top-K over the complete pairwise correlation space, computed blockwise without materializing the dense N×N result matrix. It includes the exact matrix-free complete-matrix engine, the masked pairwise-complete phosphosite analysis engine, permutation/bootstrap validation, figure-generation scripts, locked result files, integrity manifests (SHA-256), and expected-output hashes. Contents:- ptmstream/ — streaming and dense exact Top-K engine, plus the masked pairwise-complete correlation engine- scripts/ — benchmark, biological analysis, resampling, PyNetCor comparison, figure-generation, and manuscript-verification scripts- results/ — locked result tables, permutation max-null summaries (B = 1,030), bootstrap summaries (B = 510), and audit files- figures/ — Figure 1 workflow schematic and Figures 2–4- data/provenance.csv — source repository, reference, exact filename, raw-file SHA-256, accession identifiers (PDC000116, PDC000120, PDC000153), and licence information for each CPTAC cohort- MANIFEST_SHA256.txt, claims.csv, CITATION.cff, .zenodo.json, CODE_AVAILABILITY.md, and RELEASE_NOTES.md The original CPTAC phosphoproteomic matrices for colon, breast, and lung adenocarcinoma are not redistributed in this archive. They remain available from the NCI Proteomic Data Commons and LinkedOmics and are pinned by the checksums and source information in data/provenance.csv. Unrelated proprietary candidate-pruning components are not included and are not required to reproduce any manuscript result. The principal matrix-free streaming, pairwise-complete correlation, statistical validation, figure-generation, and manuscript-claim verification workflows are included. Data used in the manuscript were generated by the National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). Code in this archive is released under the MIT License. The original CPTAC source data remain subject to their original CC-BY 4.0 / NCI CPTAC terms.