Key Datasets

Datasets organized by what analyses they support. Each entry notes the phosphoproteomics platform, sample count, and what makes it uniquely useful.

Pan-Cancer Atlases

CPTAC Pan-Cancer Phosphoproteome (Geffen et al., 2023)

• Cancer types: 10 (breast, ovarian, endometrial, colon, lung adenocarcinoma, lung squamous, head and neck, glioblastoma, renal clear cell, pancreatic)
• Samples: ~1,100 tumors with matched normal
• Phosphosites: 110,274 quantified
• Platform: TMT-based DDA (Orbitrap)
• Access: CPTAC Data Portal
• Best for: Pan-cancer kinase activity analysis, phospho-subtype discovery, cross-cancer convergence studies
• Caveat: TMT ratio compression affects quantitative accuracy for small fold changes

CPTAC Proteogenomic Ecosystem (Vasaikar et al., 2023)

• Cancer types: 14 (extends Geffen with additional tumor types)
• Samples: 1,524 tumors
• Data layers: Proteomics, phosphoproteomics, genomics, transcriptomics, clinical
• Access: LinkedOmicsKB
• Best for: Multi-omic integration, proteogenomic driver discovery

Single-Cancer Datasets

Breast Cancer CPTAC (Krug et al., 2020)

• Samples: 122 treatment-naive breast tumors
• Phosphosites: ~40,000 quantified
• Key value: PAM50 subtype-specific kinase activities; foundational for phospho-subtype concept
• Best for: Subtype-specific kinase activity, benchmark dataset for method development

Ovarian Cancer CPTAC (McDermott et al., 2020)

• Samples: 169 high-grade serous ovarian tumors
• Key value: Links phospho-signaling to platinum sensitivity
• Best for: Therapeutic response prediction, platinum resistance mechanisms

Lung Adenocarcinoma CPTAC (Gillette et al., 2020)

• Samples: 110 tumors
• Key value: Connects driver mutations to phospho-signaling consequences
• Best for: Mutation-to-phospho linking, EGFR/ALK/KRAS signaling studies

Kinase Reference Datasets

KinaseLibrary Positional Scanning Data (Johnson et al., 2023)

• Kinases: 303 Ser/Thr kinases
• Method: Positional scanning peptide arrays
• Access: kinase-library.phosphosite.org
• Best for: Motif-based kinase-substrate prediction, motif comparison

PhosphoSitePlus Curated Annotations

• Phosphosites: ~240,000 human sites
• Kinase-substrate pairs: ~10,000 curated
• Access: phosphosite.org
• Best for: Training and evaluating kinase-substrate inference tools
• Caveat: Heavy bias toward well-studied kinases (AKT, ERK, CDK families); ~130 kinases have zero annotated substrates

Perturbation Datasets (for Benchmarking)

Kinase Inhibitor Panels

Multiple studies have profiled phosphoproteome changes after kinase inhibitor treatment:

Study	Inhibitors Tested	Cell Lines	Value
Wilkes et al. 2015	28 kinase inhibitors	Jurkat, A549	benchmarKIN evaluation dataset
Klaeger et al. 2017	243 clinical kinase inhibitors	Cell-free kinobeads	Drug-kinase binding ground truth
Hijazi et al. 2020	EGF/HGF time courses	HeLa	Temporal dynamics benchmark

Best for: Evaluating kinase activity inference methods; the controlled perturbation provides ground truth for which kinases should change.

Dataset Selection Guide

Analysis Goal	Recommended Dataset	Why
Pan-cancer kinase landscape	CPTAC Geffen 2023	Largest standardized phospho dataset
Method benchmarking (kinase inference)	Kinase inhibitor panels	Controlled perturbation = ground truth
Motif-based prediction	KinaseLibrary	Most comprehensive motif data
Subtype-specific analysis	Krug 2020 (breast)	Best-characterized phospho subtypes
Multi-omic integration	Vasaikar 2023 ecosystem	Matched multi-omic layers
Training ML models	PhosphoSitePlus + dbPTM	Largest labeled datasets