Dataset Information

Federated deep learning enables cancer subtyping by proteomics

ABSTRACT: Artificial intelligence (AI) applications in biomedical settings face challenges such as data privacy and regulatory compliance. Federated Deep Learning (FDL) effectively addresses these issues. We developed ProCanFDL, where local models were trained on simulated sites using proteomic data drawn from a pan-cancer cohort (n = 1,260) and 29 other cohorts (n = 6,265), representing 4,956 patients and 19,930 mass spectrometry (MS) runs, all held behind private firewalls. Local parameter updates were aggregated to build the global model, achieving a 43% performance gain over local models on the hold-out test set (n = 625) in 14 cancer subtyping tasks. Additionally, ProCanFDL preserved data privacy while matching centralized model performance. External validation assessed generalization by retraining the global model with data from two external cohorts (n = 55) and eight (n = 832) using a different MS technology. ProCanFDL presents a solution for internationally collaborative machine learning initiatives using proteomic data while maintaining data privacy.

INSTRUMENT(S): TripleTOF 6600

ORGANISM(S): Homo Sapiens (human)

TISSUE(S): Whole Body

SUBMITTER: Zainab Noor

LAB HEAD: Peter G

PROVIDER: PXD056810 | Pride | 2025-05-29

REPOSITORIES: Pride

ACCESS DATA

Dataset's files

Source:

			Action	DRS
	200817_B1-1-T1-1_00H2P_015ID_M01_S_1.wiff	Wiff
	200817_B1-1-T1-1_00H2P_015ID_M01_S_1.wiff.scan	Wiff
	200817_B1-11-T1-1_00OKG_015IN_M01_S_1.wiff	Wiff
	200817_B1-11-T1-1_00OKG_015IN_M01_S_1.wiff.scan	Wiff
	200817_B1-12-T1-1_00G59_015IO_M01_S_1.wiff	Wiff

Items per page:

1 - 5 of 7557

Similar Datasets

Project description:Patients with loss-of-function mutations of the von Hippel Lindau gene (“VHL-patients”) often develop clear cell renal cell carcinoma (ccRCC). Using archived, formalin-fixed, paraffin-embedded (FFPE) samples and a cohort of eight cases, we sought to determine global proteome alterations that distinguish ccRCC tissue from adjacent, non-malignant kidney tissue in VHL-patients. Our quantitative proteomic analysis clearly discriminated tumor and non-malignant tissue, yielding comparable proteome profiles for the eight ccRCC cases. Limma statistics was used to identify significantly dysregulated proteins in ccRCC. Functional classification of these proteins highlighted a decrease in proteins involved in the tricarboxylic acid cycle and an increase in proteins involved in glycolysis. This profile possibly represents a proteomic fingerprint of the “Warburg effect”, which is a molecular hallmark of ccRCC. Furthermore, we observed an increase in proteins involved in extracellular matrix organization. Of particular note were opposing alterations of Xaa-Pro Aminopeptidases-1 and -2 (XPNPEP-1 and -2): a strong decrease of XPNPEP-2 in ccRCC was accompanied by a strong increase of the related protease XPNPEP-1. In both cases, we corroborated the proteomic results by immunohistochemical analysis of ccRCC and adjacent, non-malignant kidney tissue of VHL patients. To functionally investigate the role of XPNPEP-1 in ccRCC, we performed small-hairpin RNA mediated XPNPEP-1 expression silencing in 786-O ccRCC cells harboring a mutated VHL gene. We found that XPNPEP-1 expression suppresses cellular proliferation and migration. These results suggest that XPNPEP-1 is rather an anti-target in VHL-driven ccRCC. Generally, we present one of the first proteomic profiling studies of ccRCC in VHL patients, comparing normal and tumor tissue, which are both deficient for the VHL tumor suppressor. Methodologically, our work further validates the robustness of using FFPE material for quantitative proteomics.

Dataset Information

Federated deep learning enables cancer subtyping by proteomics

Dataset's files

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets