Harmonization of quantitative proteomics data collected across the international cancer proteogenome consortium

The International Cancer Proteogenome Consortium (ICPC) is a collaboration among leading cancer and proteogenomics research centers. This study had the goal to assess whether different laboratories could obtain comparable quantitative proteomics data independent of the protocol, instrument platform, or analysis strategy. To perform this analysis, nine samples were derived from the NCI-7 cell line panel (Clark et al, J. Prot Res, 2018). The NCI-7 Cell Line Panel is composed of the NCI-H23, RPMI-8226, T47D, A549, COLO205, NCI-H226, and CCRF CEM cell lines which represent multiple tissues and of a variety of genetic mutations. Cell lines were grown, lysed and aliquoted at the Frederick National Laboratory for Cancer Research and sent to the University of Washington to coordinate distribution of lysates to 16 labs representing 10 countries. Each lab was sent lysates from each of the seven individual cell lines, a 4-pool lysate (NCI-H23, T47D, A549, and CCRF CEM) and the 7-pool lysate, the NCI-7 universal proteomic reference. The use of an additional reference material of each lab"s choice was encouraged. Each lab analyzed these lysates with their preferred sample preparation protocol, mass spectrometry acquisition type and instrumentation. Sample preparation protocols used in this study included S-trap, PAC/SP3, in-solution digestion (SDS or urea), in-gel digestion, and FASP. Liquid chromatography systems used include Sciex nanoLC 425, Evosep One, Thermo Easy nLC, Thermo Neo Vanquish, and Thermo UltiMate 3000. Mass spectrometers used in this analysis included Bruker timsTOF HT or Pro2, SCIEX 6600 Triple TOF, Thermo Q-Exactive (HF, HF-X, or Plus), Thermo Orbitrap Fusion Lumos, Thermo Orbitrap Eclipse, Thermo Orbitrap Fusion Lumos-FAIMS Pro, and Thermo Exploris 480. Quantitative acquisition methods were either DDA-LFQ, DDA-TMT, DIA or DIA-PASEF.

A total of 20 datasets were collected as some labs performed multiple analyses. All DDA-TMT data was analyzed using the NCI Common Data Analysis Pipeline (CDAP). All DIA data were analyzed with a DIA CDAP that makes use of either EncyclopeDIA or DIA-NN as the search engine and Skyline for quantification. The DDA-LFQ analyses were analyzed with MSAmanda and Skyline. Protein parsimony was performed using Skyline and normalized protein quantities derived using directLFQ. Batch correction was performed between each dataset using ComBat. Following normalization and batch correction, each cell line clustered together, regardless of location, sample preparation methods or instrumentation. The 4-pool lysate also clusters as expected centered amongst the 4 individual cell lines.