AUTHOR=Yu-Qing Huang , Peng-Ping Li , Ke Sun , Ke-Xing Yin , Wei-Jun Zhang , Zhen-Yu Wang TITLE=Comprehensive analysis of liquid-liquid phase separation-related genes in prediction of breast cancer prognosis JOURNAL=Frontiers in Genetics VOLUME=Volume 13 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2022.834471 DOI=10.3389/fgene.2022.834471 ISSN=1664-8021 ABSTRACT=Objective Liquid-liquid phase separation (LLPS) was a membraneless functional unit formed by specific molecules, which has been reported to play a crucial role in tumor drug resistance and growth by regulating gene expression and drug distribution. However, LLPS in prediction of cancer prognosis was not clear. This study aimed to construct prognostic model for breast based on LLPS-correlated genes. Methods LLPS-correlated genes were collected from PhaSepDB, gene expression profile and clinical characteristics of breast cancer were collected from TCGA and cBioportal, amongst which PanCancer Atlas (TCGA) cohort was as train set to construct prognostic model, and the Nature 2012&Nat Commun2016 (TCGA) cohort was as test set to perform external verification. Data analysis was performed in R and SPSS. Results We screened out 54 LLPS-related prognostic genes (p<0.01) in all samples breast cohort, 123 genes (p<0.01) in luminal breast cohort, and 28 genes (p<0.05) in triple negative (TNBC) breast cohort. 12 genes in breast cohort (training cohort: 5-year ROC=0.76, 10-year ROC=0.77; verification cohort: 5-year ROC=0.61, 10-year ROC=0.58), 8 genes in luminal cohort (training cohort: 5-year ROC=0.79, 10-year ROC=0.75; verification cohort: 5-year ROC=0.62, 10-year ROC=0.62), and 4 genes in TNBC (training cohort: 5-year ROC=0.73, 10-year ROC=0.79; verification cohort: 5-year ROC=0.55, 10-year ROC=0.54) were screened out to construct prognostic prediction model. 17-gene prognostic riskscore was constructed in all breast cohort (1-year ROC: 0.88, 3-year ROC: 0.83, 5-year ROC: 0.81), 11-gene prognostic riskscore was constructed in luminal cohort (1-year ROC: 0.67, 3-year ROC: 0.85, 5-year ROC: 0.84), and 6-gene prognostic riskscore was constructed in TNBC cohort (1-year ROC: 0.87, 3-year ROC: 0.88, 5-year ROC: 0.81). Finally, multi-gene prognostic riskscore and clinical factors were screened out to construct nomogram in all breast cohort (1-year ROC: 0.89, 3-year ROC: 0.79, 5-year ROC: 0.75, C-index=0.784), in luminal cohort (1-year ROC: 0.84, 3-year ROC: 0.83, 5-year ROC: 0.85, C-index=0.803), and in TNBC cohort (1-year ROC: 0.95, 3-year ROC: 0.84, 5-year ROC: 0.77, C-index=0.847). Conclusion This study explored the efficiency of LLPS-related genes in prediction of breast cancer prognosis, and constructed a prognosis prediction model.