CoChat

Causal integration of multi‐omics data with prior knowledge to generate mechanistic hypotheses

Open access
Dugourd, Aurelien, Kuppe, Christoph, Sciacovelli, Marco, Gjerga, Enio, Gabor, Attila, Emdal, Kristina B., Vieira, Vitor, Bekker‐Jensen, Dorte B., Kranz, Jennifer, Bindels, Eric.M.J., Costa, Ana S.H., Sousa, Abel, Beltrao, Pedro, Rocha, Miguel, Olsen, Jesper V., Frezza, Christian, Kramann, Rafael, Saez‐Rodriguez, Julio · 2021 · Molecular Systems Biology

Published version of a preprint

The DOI matches its Crossref record; 88 of the 89 references it cites were found in a registry or an index.

Abstract

Multi‐omics datasets can provide molecular insights beyond the sum of individual omics. Various tools have been recently developed to integrate such datasets, but there are limited strategies to systematically extract mechanistic hypotheses from them. Here, we present COSMOS (Causal Oriented Search of Multi‐Omics Space), a method that integrates phosphoproteomics, transcriptomics, and metabolomics datasets. COSMOS combines extensive prior knowledge of signaling, metabolic, and gene regulatory networks with computational methods to estimate activities of transcription factors and kinases as well as network‐level causal reasoning. COSMOS provides mechanistic hypotheses for experimental observations across multi‐omics datasets. We applied COSMOS to a dataset comprising transcriptomics, phosphoproteomics, and metabolomics data from healthy and cancerous tissue from eleven clear cell renal cell carcinoma (ccRCC) patients. COSMOS was able to capture relevant crosstalks within and between multiple omics layers, such as known ccRCC drug targets. We expect that our freely available method will be broadly useful to extract mechanistic insights from multi‐omics studies.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.15252/msb.20209730

Cites

88 of 88 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Overview of analysis pipeline From left to right: We sampled and processed 11 patient tumors and healthy kidney tissues from the same kidney through RNA‐sequencing and 9 of those same patients through mass spectrometry to characterize their transcriptomics, phosphoproteomics, and metabolomics profiles. We calculated differential abundance for each detected gene, phosphopeptide, and metabolite. We estimated kinase and transcription factor activities using the differential analysis statistics and footprint‐based methods. We used the estimated activities alongside the differential metabolite abundances to contextualize (i.e., extract the subnetwork that better explains the phenotype of interest) a generic trans‐omics causal prior knowledge network (meta PKN).
Figure 1 Overview of analysis pipeline From left to right: We sampled and processed 11 patient tumors and healthy kidney tissues from the same kidney through RNA‐sequencing and 9 of those same patients through mass spectrometry to characterize their transcriptomics, phosphoproteomics, and metabolomics profiles. We calculated differential abundance for each detected gene, phosphopeptide, and metabolite. We estimated kinase and transcription factor activities using the differential analysis statistics and footprint‐based methods. We used the estimated activities alongside the differential metabolite abundances to contextualize (i.e., extract the subnetwork that better explains the phenotype of interest) a generic trans‐omics causal prior knowledge network (meta PKN).CC BY · PubMed Central
Figure EV1 First two components of PCA of omic datasets of kidney tumor and healthy samples PCA of metabolomics (107 metabolites), phosphoproteomics (14,243 phosphosites) and transcriptomics (15,919 transcripts) datasets for tumor and healthy tissues samples. For each omics dataset, PCA is run independently on normalised datasets and the first two components are plotted. Each omics shows a clear separation between tumor and healthy tissue.
Figure EV1 First two components of PCA of omic datasets of kidney tumor and healthy samples PCA of metabolomics (107 metabolites), phosphoproteomics (14,243 phosphosites) and transcriptomics (15,919 transcripts) datasets for tumor and healthy tissues samples. For each omics dataset, PCA is run independently on normalised datasets and the first two components are plotted. Each omics shows a clear separation between tumor and healthy tissue.CC BY · PubMed Central
Figure EV2 Differential analysis of kidney tumor and healthy sample Combined volcano plot of metabolomics (107 metabolites), phosphoproteomics (14,243 phosphosites) and transcriptomics (15,919 transcripts), displaying the surface occupied by the points of each type of omic data. X axis represent the log2 fold change between healthy and tumour. Y axis represents the −log10 P‐value of a unpaired moderated t‐statistic obtained from using LIMMA.
Figure EV2 Differential analysis of kidney tumor and healthy sample Combined volcano plot of metabolomics (107 metabolites), phosphoproteomics (14,243 phosphosites) and transcriptomics (15,919 transcripts), displaying the surface occupied by the points of each type of omic data. X axis represent the log2 fold change between healthy and tumour. Y axis represents the −log10 P‐value of a unpaired moderated t‐statistic obtained from using LIMMA.CC BY · PubMed Central
Figure 2 Differentially regulated transcription factor, kinase, and phosphatase activities cancer vs. healthy tissue Bar plot displaying the normalized enrichment score (NES, proxy of activity change) of the 25 up‐ or down‐regulated TF and top 25 up‐ or down‐regulated kinase and phosphatases activities between kidney tumor and adjacent healthy tissue.Right panel shows the 10 most changing RNA abundances of the STAT2 regulated transcripts . Left panel shows the change of abundances of all STAT2 regulated transcripts that were used to estimate its activity change. X‐axis represents log fold change of regulated transcripts multiplied by the sign of regulation (−1 for inhibition and 1 for activation of transcription). Y‐axis represents the significance of the log fold change (−log10 of P‐value, LIMMA moderated unpaired t‐test P‐values). The black line is defined by the following function when fold change is negative : y = abs(hAss − 1 + x/(x + vAss)); and y = abs(hAss − 1 + x/(x − vAss)) when fold change is positive. abs() is the absolute value, hAss is the horizontal asymptote (hAss = 1.3) and vAss is the vertical asymptote (vAss = 0.3).Right panel shows the 10 most changing phosphopeptide abundances of the CDK7 regulated phosphopeptides. Left panel shows the change of abundances of all CDK7 regulated phosphopeptides that were used to estimate its activity change. X‐axis represents log fold change of regulated transcripts multiplied by the sign of regulation (−1 for inhibition and 1 for activation of transcription). Y‐axis represents the significance of the log fold change (−log10 of P‐value, LIMMA moderated unpaired t‐test P‐values). The black line is defined by the following function when fold change is negative : y = abs(hAss − 1 + x/(x + vAss)); and y = abs(hAss − 1 + x/(x − vAss)) when fold change is positive. Where abs() is the absolute value, hAss is the horizontal asymptote (hAss = 1.3) and vAss is the vertical asymptote (vAss = 0.3).
Figure 2 Differentially regulated transcription factor, kinase, and phosphatase activities cancer vs. healthy tissue Bar plot displaying the normalized enrichment score (NES, proxy of activity change) of the 25 up‐ or down‐regulated TF and top 25 up‐ or down‐regulated kinase and phosphatases activities between kidney tumor and adjacent healthy tissue.Right panel shows the 10 most changing RNA abundances of the STAT2 regulated transcripts . Left panel shows the change of abundances of all STAT2 regulated transcripts that were used to estimate its activity change. X‐axis represents log fold change of regulated transcripts multiplied by the sign of regulation (−1 for inhibition and 1 for activation of transcription). Y‐axis represents the significance of the log fold change (−log10 of P‐value, LIMMA moderated unpaired t‐test P‐values). The black line is defined by the following function when fold change is negative : y = abs(hAss − 1 + x/(x + vAss)); and y = abs(hAss − 1 + x/(x − vAss)) when fold change is positive. abs() is the absolute value, hAss is the horizontal asymptote (hAss = 1.3) and vAss is the vertical asymptote (vAss = 0.3).Right panel shows the 10 most changing phosphopeptide abundances of the CDK7 regulated phosphopeptides. Left panel shows the change of abundances of all CDK7 regulated phosphopeptides that were used to estimate its activity change. X‐axis represents log fold change of regulated transcripts multiplied by the sign of regulation (−1 for inhibition and 1 for activation of transcription). Y‐axis represents the significance of the log fold change (−log10 of P‐value, LIMMA moderated unpaired t‐test P‐values). The black line is defined by the following function when fold change is negative : y = abs(hAss − 1 + x/(x + vAss)); and y = abs(hAss − 1 + x/(x − vAss)) when fold change is positive. Where abs() is the absolute value, hAss is the horizontal asymptote (hAss = 1.3) and vAss is the vertical asymptote (vAss = 0.3).CC BY · PubMed Central
Figure 3 Graphical explanation of meta PKN sources A–C Schematic representation of the meta generic network (meta PKN) created combining STITCH, OmniPath and Recon3D. (A) STITCH provides information on inhibition/activation of enzyme activities mediated by metabolites. (B) OmniPath provides information inhibition/activation of enzyme activities mediated by other enzymes based mainly on curated resources. (C) Recon3D provides information on reactants and products associated with metabolic enzymes. To make this information compatible with the causal edges from OmniPath and STITCH, the interactions of recon3D are converted so that reactants “activate” their metabolic enzymes, which themselves “activate” their products.
Figure 3 Graphical explanation of meta PKN sources A–C Schematic representation of the meta generic network (meta PKN) created combining STITCH, OmniPath and Recon3D. (A) STITCH provides information on inhibition/activation of enzyme activities mediated by metabolites. (B) OmniPath provides information inhibition/activation of enzyme activities mediated by other enzymes based mainly on curated resources. (C) Recon3D provides information on reactants and products associated with metabolic enzymes. To make this information compatible with the causal edges from OmniPath and STITCH, the interactions of recon3D are converted so that reactants “activate” their metabolic enzymes, which themselves “activate” their products.CC BY · PubMed Central
Figure 4 COSMOS subnetwork centered on the interferon gamma response pathway The figure includes the main members of the interferon gamma response pathway, the most enriched cancer hallmark in the full COSMOS network. We also display the metabolic enzymes that were hypothesized to be influenced downstream of this pathway, such as BCAT1 and PNP. The numbered mechanisms are discussed in the main text.
Figure 4 COSMOS subnetwork centered on the interferon gamma response pathway The figure includes the main members of the interferon gamma response pathway, the most enriched cancer hallmark in the full COSMOS network. We also display the metabolic enzymes that were hypothesized to be influenced downstream of this pathway, such as BCAT1 and PNP. The numbered mechanisms are discussed in the main text.CC BY · PubMed Central
Figure EV3 Ranked over‐represented pathways in COSMOS solution networks A, BComparison of over‐representation analysis performed with msigDB HALLMARK pathways between COSMOS networks generated from (A) our patient samples and (B) CPTAC patient cohort. Overall, a very similar set of pathways was significantly over‐represented in both cases, notably PI3K‐AKT‐MTOR signaling, TNFA signaling via NFKB, interferon gamma response, WNT beta catenin signaling, G2M checkpoint and IL6 JAK STAT3 signaling.
Figure EV3 Ranked over‐represented pathways in COSMOS solution networks A, BComparison of over‐representation analysis performed with msigDB HALLMARK pathways between COSMOS networks generated from (A) our patient samples and (B) CPTAC patient cohort. Overall, a very similar set of pathways was significantly over‐represented in both cases, notably PI3K‐AKT‐MTOR signaling, TNFA signaling via NFKB, interferon gamma response, WNT beta catenin signaling, G2M checkpoint and IL6 JAK STAT3 signaling.CC BY · PubMed Central
Figure EV4 Overview of sample collection and study design Schematic of the process sample collection and multi‐omics data generation. We included a total of 22 samples from 11 renal cancer patients (6 men, age 65.0 ± 14.31, 5 women, age 65.2 ± 9.257(mean ± SD)) for transcriptomics. Phosphoproteomics was also measured in a subset of 18 samples from 9 of these patients (6 men, age 65 ± 14.31; 3 women, age 63.33 ± 11.06(mean ± SD)), and metabolomics was also measured in 16 samples from 8 out of these 9 patient (5 men, age 62 ± 13.23; 3 women, age 63.33 ± 9.89(mean ± SD), Dataset EV1. Patients underwent nephrectomy due to renal cancer. We processed tissue from within the cancer and a distant unaffected area of the same kidney. The tissue was snap‐frozen immediately after nephrectomy within the operation room. Figure parts created with BioRender.com.
Figure EV4 Overview of sample collection and study design Schematic of the process sample collection and multi‐omics data generation. We included a total of 22 samples from 11 renal cancer patients (6 men, age 65.0 ± 14.31, 5 women, age 65.2 ± 9.257(mean ± SD)) for transcriptomics. Phosphoproteomics was also measured in a subset of 18 samples from 9 of these patients (6 men, age 65 ± 14.31; 3 women, age 63.33 ± 11.06(mean ± SD)), and metabolomics was also measured in 16 samples from 8 out of these 9 patient (5 men, age 62 ± 13.23; 3 women, age 63.33 ± 9.89(mean ± SD), Dataset EV1. Patients underwent nephrectomy due to renal cancer. We processed tissue from within the cancer and a distant unaffected area of the same kidney. The tissue was snap‐frozen immediately after nephrectomy within the operation room. Figure parts created with BioRender.com.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

2 retracted

Show all 88 references

Cited by

Items 101–171 of 171

Metadata

Authors
Dugourd, A., Kuppe, C., Sciacovelli, M., Gjerga, E., Gabor, A., Emdal, K. B., Vieira, V., Bekker‐Jensen, D. B., Kranz, J., Bindels, E. M. J., Costa, A. S. H., Sousa, A., Beltrao, P., Rocha, M., Olsen, J. V., Frezza, C., Kramann, R., & Saez‐Rodriguez, J.
Affiliations
  • Erasmus MC (NL)
  • University of Cambridge (GB)
  • Heidelberg University (DE)
  • University of Copenhagen (DK)
  • Sankt-Antonius-Hospital Eschweiler (DE)
  • European Bioinformatics Institute (GB)
  • University of Minho (PT)
Corresponding author
  • Frezza, Christian
  • Kramann, Rafael
Venue
Molecular Systems Biology
Published
2021-01-01
Volume
17
Issue
1
Pages
e9730-e9730
ISSN
1744-4292 (ISSN-L)
OpenAlex
W3124569282
Cited by
149
References
89
Citations by year
2026: 20 · 2025: 27 · 2024: 28 · 2023: 27 · 2022: 26 · 2021: 20
FWCI
8.15
Type
journal article
Language
en
Open access
Licence
Locations
  • Molecular Systems Biology (publishedVersion, other-oa)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Research at the University of Copenhagen (University of Copenhagen) (publishedVersion, cc-by)
  • EUR Research Repository (Erasmus University Rotterdam) (publishedVersion, cc-by)
  • Cold Spring Harbor Laboratory Institutional Repository (Cold Spring Harbor Laboratory) (acceptedVersion)
  • EUR Research Repository (Erasmus University Rotterdam) (submittedVersion, cc-by)
  • RePub (Erasmus University Rotterdam) (submittedVersion, other-oa)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • RWTH Publications (RWTH Aachen) (submittedVersion, other-oa)
  • Portuguese National Funding Agency for Science, Research and Technology (RCAAP Project by FCT) (submittedVersion, cc-by)
  • Portuguese National Funding Agency for Science, Research and Technology (RCAAP Project by FCT) (submittedVersion, cc-by-nc-nd)
  • RePub (Erasmus University Rotterdam) (submittedVersion)
  • Apollo (University of Cambridge) (submittedVersion, cc-by)
  • Repository for Publications and Research Data (ETH Zurich) (submittedVersion, cc-by)
  • Apollo (cc-by)
  • RWTH Publications (RWTH Aachen)
  • ETH Zürich Research Collection (cc-by)
Funders
  • Medical Research Council (MC_UU_12022/6)
  • Novo Nordisk Fonden (NNF14CC0001)
  • Lundbeckfonden (R193‐2015‐243)
  • H2020 European Research Council (ERC‐StG 677448)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Bioinformatics and Genomic Networks
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene Regulatory Network Analysis
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods
Keywords
multi-omics integration, cosmos, phosphoproteomics, transcriptomics, metabolomics, prior knowledge incorporation, signaling networks, gene regulatory networks, metabolic networks, clear cell renal cell carcinoma, therapeutic target, cancer
MeSH terms
Carcinoma, Renal Cell, Humans, Kidney Neoplasms, Phosphoproteins, Case-Control Studies, Computational Biology, Gene Expression Profiling, Gene Regulatory Networks, Metabolomics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.