ProteoMixture: A Cell Type Deconvolution Tool for Bulk Tissue Proteomics Data

Abstract

Numerous multi-omic investigations of cancer tissue have documented varying and poor pairwise transcript:protein quantitative correlations and most deconvolution tools aiming to predict cell type proportions (cell admixture) have been developed and credentialed using transcript-level data alone. To estimate cell admixture using protein abundance data, we analyzed proteome and transcriptome data generated from contrived admixtures of tumor, stroma, and immune cell models or those selectively harvested from the tissue microenvironment by laser microdissection from high grade serous ovarian cancer (HGSOC) tumors. Co-quantified transcripts and proteins performed similarly to estimate stroma and immune cell admixture in two commonly used deconvolution algorithms ESTIMATE and ConsensusTME (r ≥ 0.63). Here we have developed and optimized protein-based signatures to estimate cell admixture proportions and benchmarked these using bulk tumor proteomics data from over 150 HGSOC patients. The optimized protein signatures supporting cell type proportion estimates from bulk tissue proteomic data are available at https://lmdomics.org/ProteoMixture/.

Mapping three-dimensional intratumor proteomic heterogeneity in uterine serous carcinoma by multiregion microsampling

“…We also explored a single sample gene set enrichment analysis (ssGSEA) classifier enabling prediction of tumor, stroma, and immune cell admixture optimized for proteomic data (ProteoMixture [ 25 ]) for characterizing the ET, ES, and BT samples (Additional file 8 : Table S8). BT samples with high tumor cellularity and ET had higher ssGSEA “tumor” scores, versus BT samples with low tumor cellularity and ES had higher ssGSEA “stroma” scores.…”

“…Of the 455 total significant protein alterations (collectively from Additional file 8 : Tables S16 and S18) between in ET versus ES in our dataset, 422 were co-quantified in the CPTAC tumors and were highly correlated (Spearman Rho = 0.93, p < 0.001). Analysis of the CPTAC proteomic data using ProteoMixture [ 25 ] and further comparison of the ProteoMixture stromal scores with the inferred stromal purity scores revealed high correlation (Spearman Rho = 0.8, p < 0.01) of CPTAC purity values with the ProteoMixture stromal score (Fig. 7 B).…”

“…p < 0.05) between ET and ES for n = 9 USC tumors (from Additional file 8 : Tables S16 & S18; n = 455 proteins total) were correlated with global proteome data from bulk tissues collections for n = 9 USC tumors reported by Dou Y et al Among 7908 total proteins quantified in all samples by Dou Y et al, 442 proteins were co-quantified in our cohort with tumor and stroma alterations. Correlation of the quantitative abundances for these 442 proteins were directly compared to metrics of tumor purity derived from analysis of methylation data (Purity_Cancer) as reported in Dou Y et al B Stroma scores were calculated for global proteome data reported by Dou Y et al using ProteoMixture [ 25 ] and directly compared to stroma purity metrics derived from methylation data (Purity_Stroma) reported by Dou Y et al …”

“…Single-sample gene set enrichment analysis (ssGSEA) [ 23 ] was performed using the protein-level data using the GSVA [ 24 ] package (version 1.34.0) in R (version 3.6.0). Signatures used for ssGSEA were generated by incorporating differential expression analysis with support vector machine recursive feature elimination (RFE) on quantitative proteomic data from a linearly diluted series of in situ-derived HGSOC ET, ES, and immune cells [ 25 ]. The ssGSEA clustermap and boxplots were plotted in Python (version 3.9.16) with seaborn (version 0.11.2), matplotlib (version 3.7.0) [ 26 ], and stat annotations (version 0.5.0) [ 27 ].…”

“…Quantitative abundances for these 442 proteins were directly compared to the methylation-derived metrics of tumor purity (Purity_Cancer), as reported in Additional file 8 : Table S1 from Dou Y et al [ 28 ]. Stroma scores for the global proteome data from the CPTAC dataset were calculated using ProteoMixture [ 25 ] and directly compared to the methylation-derived stroma purity metrics (Purity_Stroma) from the CPTAC dataset [ 28 ].…”