Skip to main content
NCBI home page
Scientific Data logoLink to Scientific Data
. 2026 Jun 29;13:1470. doi: 10.1038/s41597-026-07756-5

A dataset supporting Combinatorial Proteome Integral Solubility/Stability Alteration Analysis (CoPISA)

Ehsan Zangene 1,#, Elham Gholizadeh 1,#, Uladzislau Vadadokhau 1, Danilo Ritz 2, Amir A Saei 3, Mohieddin Jafari 1,4,5,✉
PMCID: PMC13687023  PMID: 42374105

Abstract

Combination therapies are widely used in acute myeloid leukemia (AML), but systematic datasets capturing proteome-wide responses to multi-drug perturbations remain limited. Here we present CoPISA (Combinatorial Proteome Integral Solubility/Stability Alteration), a quantitative proteomics assay designed to profile protein solubility and stability responses to single and combined drug treatments. The dataset includes two AML drug pairs (LY3009120–sapanisertib and ruxolitinib–ulixertinib) applied to four AML cell lines (MOLM-13, MOLM-16, SKM-1, and NOMO-1) under control, single-agent, and combination conditions in both lysate and intact-cell formats. Thermal solubility profiling coupled with TMT-based multiplexed LC–MS/MS generated 16 TMT16-plex experiments comprising 192 LC–MS/MS raw files, providing deep proteome coverage across treatments and biological contexts. The resource includes raw and processed proteomics data, detailed experimental metadata in Sample and Data Relationship Format (SDRF), and reproducible analysis scripts for reporter normalization, protein-level aggregation, statistical modeling, and classification of combinatorial response patterns. The experimental design enables identification of proteins responding uniquely to combination treatments as well as overlapping single-agent effects. Technical validation demonstrates reproducible quantification across multiplex experiments and assay formats. All data are publicly available through the PRIDE repository (PXD066812) together with analysis code, enabling independent reanalysis and method development. This dataset provides a benchmark resource for studying proteome responses to drug combinations, comparing lysate and intact-cell perturbation profiles, developing computational approaches for combinatorial target inference, and supporting training in computational proteomics.

Background & Summary

Combination therapy is a cornerstone of AML treatment, yet systematically capturing how multiple drugs jointly reconfigure the proteome remains challenging1,2. Traditional synergy metrics quantify abundance-level phenotypic effects of drug pairs but do not reveal underlying protein-level responses3 or combination-specific mechanisms4–6. To address this, we developed the Combinatorial Proteome Integral Solubility/Stability Alteration (CoPISA) assay, a high-throughput proteomics approach that measures protein solubility and stability shifts under single- and multi-drug perturbations. CoPISA provides a readout of protein engagement in lysate and downstream effects in intact cells, enabling the detection of effects that emerge only under combination treatments.

The dataset encompasses two rational AML drug pairs7 (i.e., LY3009120–sapanisertib (LS) and ruxolitinib–ulixertinib (RU)) applied to four AML cell lines (i.e., NOMO-1, MOLM-13, MOLM-16, and SKM-1) under control, individual drug, and combination conditions in both cell lysate and intact-cell contexts. Protein responses were classified according to whether solubility shifts occur under a single agent, both agents, or exclusively under the combination. This framework operationalizes combination responses using logical-gate concepts8. An “AND-gate” or conjunctional pattern denotes proteins that change significantly only under the combination treatment (A + B) but not under either single agent (A or B), whereas an “OR-gate” or disjunctional pattern denotes proteins responding to at least one single agent and also present in the combination condition (Fig. 1). These definitions enable systematic separation of combination-specific and overlapping proteome responses9.

Fig. 1.

Fig. 1

Open in a new tab

Identification of combination-specific protein responses. Proteomic profiles are compared across three treatment conditions: Drug A, Drug B, and the combination treatment (A + B). Each condition reveals a set of detected proteins, represented as individual protein structures. While some proteins respond to single treatments (A or B), a subset of proteins becomes significantly altered only in the presence of the combination therapy. These proteins are absent or non-significant in both single-agent conditions but emerge under the combined perturbations. The highlighted region illustrates proteins uniquely detected in the A + B condition, representing combination-specific responses consistent with an AND-gate behavior, where a significant effect occurs only when both treatments are present simultaneously.

To interpret these protein responses quantitatively, we classified proteins into logical response categories using statistical significance thresholds applied independently to each treatment contrast. Inspired by Boolean logic used in computer science, we defined two response patterns. An “AND-gate” response refers to proteins that become significantly altered only when both drugs are applied together while remaining non-significant in either single-agent treatment. This pattern highlights proteins whose response emerges specifically from the combined perturbation, suggesting cooperative or synergistic effects between the drugs. In contrast, an “OR-gate” response refers to proteins that are significantly altered in at least one single-drug treatment and are also detected in the combination condition. These proteins therefore reflect shared or overlapping drug effects rather than combination-specific responses. This logic-based framework enables systematic separation of proteins uniquely driven by drug combinations from those responding to individual drugs (Fig. 1).

This logic-based framework enables systematic separation of proteins uniquely driven by drug combinations from those responding to individual drugs. To support reuse of these results, we also provide an interactive Shiny web application where readers can search, filter, visualize, and download the significant protein responses and their AND-/OR-gate classifications across drug conditions, cell lines, and sample contexts: https://ehsan-zangene.shinyapps.io/Copisa_Guid/.

The resource includes raw and processed multiplexed proteomics data, differential solubility outputs for each treatment and context, intersection analyses, and refined post-translational modification (PTM)-aware identifications. Across four AML cell lines, two assay formats (lysate and intact-cell), four treatment conditions (control, single agent 1 and 2, and combinations), and three technical replicates per condition, the dataset comprises 16 TMT16-plex acquisitions (192 LC–MS/MS raw files). Technical replicates were generated by splitting each sample into three independent aliquots, each subjected to separate PISA processing and labeled with distinct TMT tags, while four AML cell lines served as biological replicates. On average, each multiplex experiment yielded several thousands of confidently quantified proteins, providing deep proteome coverage suitable for comparative solubility profiling. The inclusion of pooled reference channels in every TMT set enables within- and cross-plex normalization, and replicate-level concordance supports quantitative reproducibility across biological contexts and treatment modalities. To facilitate automated reprocessing, we provide an SDRF file that encodes sample relationships and experimental metadata10,11. Data collection is summarized in Fig. 2.

Fig. 2.

Fig. 2

Open in a new tab

Experimental design and biological and technical replication strategy for CoPISA assay. The study included four AML cell lines analyzed using two assay formats (lysate/extract and intact/live-cell) under eight treatment conditions (control, drug A, drug B, and combination for both RU and LS combinations), each measured in three technical replicates. For each biological context (cell line × assay format), samples were organized into independent TMTpro multiplexes containing 12 biological channels and one pooled reference channel. Each multiplex was fractionated into 12 concatenated high-pH fractions and analyzed by LC–MS/MS, resulting in 192 raw LC–MS/MS files and 2,496 quantified signals across the complete study. One sample/channel was excluded during quality control due to poor labeling efficiency, resulting in 191 samples used for downstream peptide and protein quantification.

These comprehensive data layers support multiple reuse scenarios, including benchmarking computational models for combinatorial target inference, exploring differences between lysate and intact-cell responses, and testing intersection-based analysis strategies for detecting combination-specific protein effects. Cleanly structured and well-annotated perturbation datasets can also serve as reusable inputs for downstream computational frameworks aimed at therapeutic hypothesis generation, including approaches that prioritize candidate vulnerabilities or combination strategies in cancer and other diseases12,13. Furthermore, the open accessibility of both the dataset and the accompanying analysis code makes this resource valuable not only for educational purposes, such as training workshops and teaching modules in computational proteomics, but also for incorporation into public-facing resources and reusable data portals that help other researchers explore, compare, and build upon solubility- and stability-based drug-response datasets14–17.

Methods

Cell culture

Human AML cell lines MOLM-13, MOLM-16, SKM-1, and NOMO-1 were obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ, Germany). Cells were maintained in RPMI-1640 medium (Gibco, Thermo Fisher Scientific, USA) supplemented with 10–20% fetal bovine serum (FBS; Gibco, Thermo Fisher Scientific), 2 mM L-glutamine (Lonza), and 100 U/mL penicillin/streptomycin (Gibco, Thermo Fisher Scientific) at 37 °C in a humidified incubator with 5% CO₂. Cells were collected by centrifugation at 400 × g for 4 min. Mycoplasma contamination was excluded using PCR-based testing.

Drug treatment conditions and thermal profiling strategy

To generate protein solubility/stability response profiles, cells or lysates were exposed to defined single agents and combinations at fixed concentrations selected to produce measurable pathway engagement while maintaining compatibility with the CoPISA thermal workflow. The following final concentrations were used: LY3009120 (500 nM), sapanisertib (500 nM), ruxolitinib (3 µM), and ulixertinib (3 µM). Vehicle controls contained 0.1% DMSO. Drug concentrations were selected based on prior optimization and validation studies of these AML drug combinations, aiming to achieve robust pathway engagement and reproducible proteome-wide solubility responses while maintaining compatibility with the CoPISA thermal profiling workflow5,7. Each sample was distributed across a temperature series spanning 48–59 °C in 1 °C increments, with a 3 min heat exposure per temperature point and a subsequent 3 min equilibration at room temperature prior to soluble fraction recovery. For each protein, the CoPISA signal per condition was defined as the log2-transformed TMT reporter intensity normalized to the pooled reference channel within each multiplex. Differential solubility (ΔS) was computed as the difference between log2-normalized intensities of treatment and matched DMSO control within the same cell line, assay mode, and multiplex. A “one-pot” pooling strategy was used per treatment condition by combining the soluble material across all temperature points after heating, thereby encoding the area under the melting curve as a single abundance-like measure per protein per condition, enabling high-throughput replication within TMT multiplexing9.

Lysate-based CoPISA sample preparation

For lysate CoPISA, 2 × 10⁶ cells per condition were collected, washed twice in PBS, and resuspended in PBS containing protease inhibitors (Roche). Cells were lysed by four freeze–thaw cycles (liquid nitrogen freeze followed by thaw at 35 °C), and insoluble debris was removed by centrifugation at 10,000 g for 10 min at 4 °C. Clarified lysates were treated with compounds for 15 min at room temperature, split into aliquots for the temperature series, heated, equilibrated, pooled per condition, and ultracentrifuged at 100,000 g for 20 min at 4 °C to isolate the soluble protein fraction used for proteomics.

Intact-cell CoPISA sample preparation

For intact-cell CoPISA, 2 × 10⁶ cells per condition were treated in culture for 60 min at 37 °C and 5% CO₂, washed twice with PBS, and resuspended in PBS with protease inhibitors. Samples were heated across the same temperature series used for lysates. After equilibration, cells were lysed using four freeze–thaw cycles, pooled per condition, and ultracentrifuged at 100,000 g for 20 min at 4 °C to recover soluble proteins. Soluble fractions were snap frozen and stored at −80 °C until processing to minimize freeze–thaw variability.

Sample Preparation and TMT Labeling

Protein concentrations were determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). For each sample, 27 µg of protein was processed using 10 kDa centrifugal filter units (Amicon Ultra-0.5 mL, 10KD) to ensure consistent buffer exchange and digestion conditions across the dataset. Samples were adjusted with 20 mM EPPS buffer, reduced with 10 mM dithiothreitol (55 °C, 30 min), and alkylated with 50 mM iodoacetamide (25 °C, 60 min, in the dark). Proteins were digested overnight at 37 °C using trypsin (Trypsin/P specificity; enzyme:protein ratio 1:50). Peptides were recovered by centrifugation and additional EPPS rinses.

To support cross-channel comparability within multiplexes, a pooled reference channel was created for each TMT set by combining equal peptide amounts from all samples included in that set. Peptides were labeled using TMTpro 16-plex reagents (Thermo Fisher Scientific) with 100 µg TMTpro reagent per 25 µg peptide input per channel, corresponding to an approximately 4:1 reagent-to-peptide mass ratio. Labeled channels were combined per TMT set into a single mixture, yielding 325 µg total peptide for each multiplexed experiment (13 channels × 25 µg per channel, including the pooled reference). Across the entire dataset, 16 independent TMT16-plex acquisitions were collected to cover the full matrix of cell lines, assay modes (lysate vs. intact), and treatment conditions.

Peptide fractionation and LC–MS/MS acquisition parameters

TMT-labeled peptide mixtures were fractionated by high-pH reversed-phase chromatography (Ultimate 3000, Thermo Fisher Scientific) using a gradient from 2% to 15% buffer B over 3 min, to 45% B over 59 min, to 80% B over 3 min with a 9 min hold, followed by re-equilibration to 2% B. Fractions were concatenated to 12 final fractions per TMT set and analyzed independently by LC–MS/MS.

LC–MS/MS was performed on an Orbitrap Eclipse Tribrid mass spectrometer coupled to an Ultimate 3000 nanoLC system with a FAIMS Pro interface (Thermo Fisher Scientific). Peptides were separated on an in-house packed C18 column at nanoflow rates. Data were acquired in data-dependent mode with SPS-MS3 quantification to improve TMT reporter accuracy. MS1 spectra were collected in the Orbitrap, MS2 in the ion trap, and MS3 in the Orbitrap. FAIMS compensation voltages of −40 V and −70 V were alternated. Dynamic exclusion was enabled. Real-time search was performed using a human UniProt database (20,362 entries; download date: 2020-04-17) to enable rapid peptide identification and more efficient SPS-MS3 selection in complex multiplexed samples.

Data Records

Raw files were processed in MaxQuant 2.6.5.0 with the Andromeda search engine against the reviewed human Swiss-Prot UniProt database (downloaded 2024-10-10), using a target-decoy strategy and common contaminants. Searches were run in Reporter MS3 mode with manufacturer-provided TMT label correction factors and weighted-ratio normalization to the reference channel. Variable modifications included Met oxidation and protein N-terminal acetylation; Cys carbamidomethylation was fixed. Up to 3 variable modifications and 2 trypsin/P missed cleavages were allowed per peptide. The refined search used the same settings but added PEIMAN2-suggested variable modifications (up to 5 per peptide)18,19. Precursor mass tolerances were 20 ppm for the first search and 4.5 ppm for the main search, with internal recalibration. Peptides ≥7 amino acids and charges up to + 7 were considered. TMT intensities were quantified in MS3 mode with isotope-impurity correction and normalized to the pooled reference. For batch-effect quality control, uncorrected reporter intensities were retained and compared with pooled-reference corrected reporter intensities and the final median-centered protein-level matrix. This enabled direct assessment of run-associated and plex-associated structure before and after pooled-reference correction. FDR was controlled at 1% at the PSM, peptide, protein, and site levels. Proteins require at least one Razor or unique peptide.

The mass spectrometry proteomics data and corresponding metadata have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD06681220. The raw data is organized in 16 archived files, corresponding to 16 TMT-labelled samples, and named as Kit i_j, where i is the TMT-labelling kit number and j is the ordinal number of the sample that was labelled with Kit i. Every archive file contains 12 raw mass spectrometry data, corresponding to 12 fractions obtained from 16 samples. In total, there are 192 raw files. Moreover, raw files were converted to the HUPO proteomics standard mzML. Unprocessed MaxQuant searches were deposited as 2 archived files. To allow the community to reproduce the MaxQuant search, we also deposited parameter files for the search software, as well as the protein database that was used in our search. Lastly, the SDRF file was also published in order to provide transparency on the experimental design and the raw files16.

Technical Validation

To support reuse of this dataset for independent reanalysis and method development, we performed technical validation21 at three levels: (i) biological input quality, (ii) experimental reproducibility of the CoPISA perturbation workflow in both extract and intact formats, and (iii) analytical performance of TMT-based multiplexed LC–MS/MS. The final processed dataset comprised 42,083 filtered peptides, which were aggregated into 5,559 proteins across 191 samples, following exclusion of one sample due to poor labeling efficiency (originally 192 samples), providing broad proteome coverage and enabling proteome-wide assessment of drug-induced solubility changes across diverse cellular pathways. Quantitative precision across biological replicates was assessed using the coefficient of variation (CV) of protein intensities within experimental conditions. Median protein CV across replicate groups was 12.32% (interquartile range, 10.91–15.61%), supporting the robustness of the multiplexed quantitative measurements.

All validation steps were designed to identify and remove problematic runs, confirm consistency of multiplexed quantification, and demonstrate that combination treatments generate non-trivial proteome-wide solubility signatures rather than artifacts driven by acquisition or processing variability. All AML cell lines (MOLM-13, MOLM-16, SKM-1, and NOMO-1) were obtained from a major biological resource center (DSMZ) and maintained under standardized culture conditions to minimize variability attributable to media composition, density, or handling.

CoPISA profiling was performed in two complementary modes, lysate-based (extract) and living-cell (intact), using the same drug concentrations, thermal challenge design, and downstream protein isolation strategy. The dual-format design provides an internal validation axis: extract emphasizes direct physicochemical effects consistent with primary drug–protein engagement in lysates, whereas intact captures integrated cellular responses that include downstream remodeling22,23. For each cell line and condition, three technical replicates were acquired, enabling replicate-level concordance assessment within the same biological background. Temperature challenges were executed using a fixed program (3 min heating followed by 3 min equilibration) across 48–59 °C in 1 °C increments, and solubility readouts were derived from identically pooled temperature-series material per treatment. The use of identical heating grids, pooling strategy, and ultracentrifugation conditions across all samples constrains technical degrees of freedom and supports comparability across treatments, cell lines, and preparation modes.

In addition to replicate-aware experimental design, run-level quality control was assessed directly from reporter ion intensity distributions across TMT channels and multiplex sets. As shown in Fig. 3, log2-transformed reporter intensity densities and boxplots were broadly concordant across TMT labels, runs, and multiplexes, indicating stable signal behavior and absence of systematic channel- or run-specific distortion. These visualizations complement replicate-level reproducibility, confirming that the multiplexed acquisition was technically consistent across the dataset before downstream normalization and protein-level inference.

Fig. 3.

Fig. 3

Open in a new tab

Quality-control assessment across TMT labels and multiplex runs. (A) Density distributions of log2-transformed reporter ion intensities stratified by TMT label. (B) Boxplots of log2-transformed reporter ion intensities across TMT labels. (C) Density distributions of log2-transformed reporter ion intensities stratified by TMT run/kit. (D) Boxplots of log2-transformed reporter ion intensities across TMT runs/kits. In the boxplots, the center line denotes the median, the box boundaries represent the first and third quartiles, and the whiskers extend to 1.5 times the interquartile range.

To control variability introduced during proteomic processing, total protein input was standardized by BCA quantification before digestion, and equal peptide amounts were used for TMT labeling. Sample processing employed established reduction/alkylation conditions and filter-aided cleanup before overnight trypsin digestion, which reduces interference from detergents and low-molecular contaminants and improves digestion reproducibility. TMT labeling was performed using consistent peptide input per channel, and each multiplex set included a pooled reference channel generated from equal peptide amounts across all samples in that set. The reference channel supports within-plex normalization and enables cross-sample comparison by anchoring reporter intensities to a common internal standard. Reporter ion intensities were normalized in two stages. First, within-plex normalization was performed by dividing each channel by the pooled reference channel to correct for loading differences. Second, between-plex normalization was applied using median centering of protein-level log2 ratios across multiplexes to minimize batch-specific shifts while preserving biological contrasts. We evaluated the effectiveness of this normalization strategy by comparing protein-level matrices generated from uncorrected reporter intensities and pooled-reference corrected reporter intensities. PCA and relative log expression diagnostics were calculated for the 191 retained biological channels and 5,638 proteins passing QC filters. Pooled-reference correction reduced plex-associated structure, with the weighted coefficient of determination (R²) calculated on principal components (PCs) for plex decreasing from 0.312 before correction to 0.243 after pooled-reference correction; the subsequent median-centering step further reduced this value to 0.123. Kit-associated weighted PC R² similarly decreased from 0.065 before correction to 0.045 after pooled-reference correction and 0.020 after median-centering. Corrected pooled reference channels were anchored as expected, with non-zero corrected pooled-channel values equal to 1. These results indicate that pooled-reference correction reduced run-associated variation while avoiding the assumption that batch effects were completely absent. Fractionation by high-pH reversed-phase chromatography was performed using a fixed gradient and concatenation scheme to yield 12 fractions per multiplex set, increasing proteome coverage and reducing ratio compression by distributing precursor complexity across runs.

To further document analytical consistency after preprocessing, we summarized sample-level normalized intensity profiles and protein-level coverage across retained channels. After filtering reverse hits, contaminants, pooled or excluded channels, features with insufficient peptide support, and non-quantified entries, peptide-level measurements were median-centered before aggregation. The resulting protein-level matrix preserved broad sample comparability while maintaining biologically meaningful treatment contrasts, supporting its suitability for downstream differential solubility modeling and method benchmarking.

All TMT fractions were analyzed on the same LC–MS/MS platform (Orbitrap Eclipse Tribrid with FAIMS Pro), using SPS-MS3 acquisition to improve quantitative accuracy of reporter ions in multiplexed experiments. Alternating FAIMS compensation voltages (−40 V and −70 V) were used to enhance depth and reduce chemical noise. Because each TMT set contains an internal pooled reference channel, channel-wise reporter distributions can be inspected for gross outliers indicative of labeling failures, injection anomalies, or fraction-level losses. Statistical inference of differential solubility was performed at the protein level using linear modeling across replicate-level log2-normalized reporter intensities. For each cell line and assay mode, contrasts were defined comparing single agents and combinations to their matched DMSO controls. Empirical Bayes moderation was applied to variance estimates to improve stability across the proteome. In addition, several orthogonal QC summaries were used to evaluate quantitative consistency, including channel-level intensity distributions, run/kit-level signal concordance, feature missingness as a function of abundance24, per-sample protein coverage, and sample-to-sample correlation structure (Fig. 3). The repeated appearance of technical replicates within the same multiplex set further enables detection of run- or fraction-specific deviations.

Usage Notes

The dataset associated with ProteomeXchange accession PXD066812 can be used to investigate protein-level responses to combinatorial therapies in acute myeloid leukemia (AML). The data support downstream applications, including drug-target interaction analysis, pathway and network analysis25,26, comparative proteomics27, biomarker discovery, and benchmarking of computational proteomics workflows16.

The repository includes RAW files, mzML files, SDRF annotation files, processed outputs, and analysis scripts to facilitate reproducibility and reuse. Researchers may reanalyze the RAW data using their preferred proteomics search software and customized search parameters. Parameter files used in this study are also provided to support the reproduction of the original analysis workflow.

The preprocessing and downstream analysis scripts for MaxQuant-based analyses are available in the Code Availability section. Researchers using alternative proteomics software may directly use the provided RAW or mzML files together with the SDRF metadata to define the experimental design and perform customized analyses.

Author contributions

E.Z. was responsible for the initial manuscript drafting, computational analyses, data interpretation, data visualization, and preparation of scientific illustrations. E.G. conducted the experimental analyses and contributed to data interpretation. U.V. participated in the proteomics data analysis, while D.R. performed the proteomics mass spectrometry experiments and associated data analysis. M.J. conceived the study, formulated the research questions, provided scientific direction, and supervised the project. All authors contributed to the preparation of the manuscript.

Funding

This study was financially supported by the Tampere Institute for Advanced Study, the Research Council of Finland [Grant 332454 to M.J.], and the Jane and Aatos Erkko Foundation [Grant 220031 to M.J.]. EZ’s salary is partially supported by the iCANPOD postdoctoral program, which is funded through the iCANDOC doctoral education pilot in precision cancer medicine.

Data availability

The mass spectrometry proteomics data and corresponding metadata are available at the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD066812 (https://www.ebi.ac.uk/pride/archive/projects/PXD066812).

Code availability

The R code used for reporter normalization, protein-level aggregation, statistical modeling, logical gate classification (AND/OR patterns), and visualization is available at the GitLab repository (https://version.helsinki.fi/ehzange/Max_RULS). The repository includes reproducible scripts for data import, quality control diagnostics, normalization workflows, linear modeling, multiple testing correction, and generation of intersectional solubility maps. All analyses were performed using R (version 4.5.2) with documented package dependencies to ensure computational reproducibility. To facilitate navigation of the repository structure and access to generated outputs, we additionally provide an interactive Shiny-based repository navigator that organizes scripts, figures, processed outputs, and documentation through dynamically generated repository links, which is available on https://ehsan-zangene.shinyapps.io/Copisa_Guid.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Ehsan Zangene, Elham Gholizadeh.

References

  • 1.Cherry, E. M. et al. Venetoclax and azacitidine compared with induction chemotherapy for newly diagnosed patients with acute myeloid leukemia. Blood Adv.5, 5565–5573 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Barneh, F. et al. Integrated use of bioinformatic resources reveals that co-targeting of histone deacetylases, IKBK and SRC inhibits epithelial-mesenchymal transition in cancer. Brief Bioinform20, 717–731 (2019). [DOI] [PubMed] [Google Scholar]
  • 3.Gholizadeh, E. et al. Identification of Celecoxib-Targeted Proteins Using Label-Free Thermal Proteome Profiling on Rat Hippocampus. Mol Pharmacol99, 308–318 (2021). [DOI] [PubMed] [Google Scholar]
  • 4.Tyner, J. et al. Functional genomic landscape of acute myeloid leukemia. Nature562, 526–531 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Mirzaie, M. et al. Designing patient-oriented combination therapies for acute myeloid leukemia based on efficacy/toxicity integration and bipartite network modeling. Oncogenesis13, 11 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Barneh, F. et al. Valproic acid inhibits the protective effects of stromal cells against chemotherapy in breast cancer: Insights from proteomics and systems biology. J Cell Biochem119, 9270–9283 (2018). [DOI] [PubMed] [Google Scholar]
  • 7.Gholizadeh, E. et al. Targeting acute myeloid leukemia resistance with two novel combinations demonstrate superior efficacy in TP53, HLA-B, MUC4 and FLT3 mutations. Biomedicine & Pharmacotherapy192, 118647 (2025). [DOI] [PubMed] [Google Scholar]
  • 8.Jafari, M., Ansari-Pour, N., Azimzadeh, S. & Mirzaie, M. A logic-based dynamic modeling approach to explicate the evolution of the central dogma of molecular biology. PLoS One12, e0189922 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Gholizadeh, E. et al. Solubility based mechanistic profiling of combinatorial drug therapy. Nat. Commun. 17, (2026). [DOI] [PMC free article] [PubMed]
  • 10.Dai, C. et al. A proteomics sample metadata representation for multiomics integration and big data analysis. Nat. Commun.12, 5854 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Claeys, T. et al. lesSDRF is more: maximizing the value of proteomics data through streamlined metadata annotation. Nat. Commun.14, 6743 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Zangene, E., Marashi, S.-A. & Montazeri, H. SL-scan identifies synthetic lethal interactions in cancer using metabolic networks. Sci Rep13, 15763 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Montazeri, H. et al. Systematic identification of novel cancer genes through analysis of deep shRNA perturbation screens. Nucleic Acids Res.49, 8488–8504 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Webel, H., Perez-Riverol, Y., Nielsen, A. B. & Rasmussen, S. Mass spectrometry-based proteomics data from thousands of HeLa control samples. Scientific Data11, 112 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Ping, L. et al. Global quantitative analysis of the human brain proteome and phosphoproteome in Alzheimer’s disease. Scientific Data7, 315 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Vadadokhau, U. et al. Preventing Proteomics Data Tombs Through Collective Responsibility and Community Engagement. Sci Data13, (2026). [DOI] [PMC free article] [PubMed]
  • 17.Zangene E, et al. DORSSAA: Drug-target interactOmics Resource based on Stability/Solubility Alteration Assay. Mol Cell Proteomics. 101603, 10.1016/j.mcpro.2026.101603 (2026). [DOI] [PMC free article] [PubMed]
  • 18.Nickchi, P. et al. Monitoring Functional Posttranslational Modifications Using a Data-Driven Proteome Informatic Pipeline. Proteomics25, e202400238 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Nickchi, P., Jafari, M. & Kalantari, S. PEIMAN 1.0: Post-translational modification Enrichment, Integration and Matching ANalysis. Database (Oxford)2015, bav037 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.PRIDE - PRoteomics IDEntifications Database. https://www.ebi.ac.uk/pride/archive/projects/PXD066812 (2026). [DOI] [PubMed]
  • 21.Jafari, M., Guan, Y., Wedge, D. C. & Ansari-Pour, N. Re-evaluating experimental validation in the Big Data Era: a conceptual argument. Genome Biol22, 71 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Martinez Molina, D. et al. Monitoring drug target engagement in cells and tissues using the cellular thermal shift assay. Science341, 84–87 (2013). [DOI] [PubMed] [Google Scholar]
  • 23.Gaetani, M. et al. Proteome Integral Solubility Alteration: A High-Throughput Proteomics Assay for Target Deconvolution. J Proteome Res18, 4027–4037 (2019). [DOI] [PubMed] [Google Scholar]
  • 24.Zangene, E., Schwämmle, V. & Jafari, M. Informative Missingness in Nominal Data: A Graph-Theoretic Approach to Revealing Hidden Structure. Comput Struct Biotechnol J35, 0099 (2026). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Jafari, M., Sadeghi, M., Mirzaie, M., Marashi, S.-A. & Rezaei-Tavirani, M. Evolutionarily conserved motifs and modules in mitochondrial protein-protein interaction networks. Mitochondrion13, 668–675 (2013). [DOI] [PubMed] [Google Scholar]
  • 26.Ansari-Pour, N., Razaghi-Moghadam, Z., Barneh, F. & Jafari, M. Testis-Specific Y-Centric Protein-Protein Interaction Network Provides Clues to the Etiology of Severe Spermatogenic Failure. J Proteome Res15, 1011–1022 (2016). [DOI] [PubMed] [Google Scholar]
  • 27.Jafari, M., Mirzaie, M., Sadeghi, M., Marashi, S.-A. & Rezaei-Tavirani, M. Exploring biological processes involved in embryonic stem cell differentiation by analyzing proteomic data. Biochim Biophys Acta1834, 1063–1069 (2013). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The mass spectrometry proteomics data and corresponding metadata are available at the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD066812 (https://www.ebi.ac.uk/pride/archive/projects/PXD066812).

The R code used for reporter normalization, protein-level aggregation, statistical modeling, logical gate classification (AND/OR patterns), and visualization is available at the GitLab repository (https://version.helsinki.fi/ehzange/Max_RULS). The repository includes reproducible scripts for data import, quality control diagnostics, normalization workflows, linear modeling, multiple testing correction, and generation of intersectional solubility maps. All analyses were performed using R (version 4.5.2) with documented package dependencies to ensure computational reproducibility. To facilitate navigation of the repository structure and access to generated outputs, we additionally provide an interactive Shiny-based repository navigator that organizes scripts, figures, processed outputs, and documentation through dynamically generated repository links, which is available on https://ehsan-zangene.shinyapps.io/Copisa_Guid.

Articles from Scientific Data are provided here courtesy of Nature Publishing Group

RESOURCES