Predictive, preventive, and personalized medicine (PPPM) is the hot spot and future direction in the field of cancer. Cancer is a complex, whole-body disease that involved multi-factors, multi-processes, and multi-consequences. A series of molecular alterations at different levels of genes (genome), RNAs (transcriptome), proteins (proteome), peptides (peptidome), metabolites (metabolome), and imaging characteristics (radiome) that resulted from exogenous and endogenous carcinogens are involved in tumorigenesis and mutually associate and function in a network system, thus determines the difficulty in the use of a single molecule as biomarker for personalized prediction, prevention, diagnosis, and treatment for cancer. A key molecule-panel is necessary for accurate PPPM practice. Pattern recognition is an effective methodology to discover key molecule-panel for cancer. The modern omics, computation biology, and systems biology technologies lead to the possibility in recognizing really reliable molecular pattern for PPPM practice in cancer. The present article reviewed the pathophysiological basis, methodology, and perspective usages of pattern recognition for PPPM in cancer so that our previous opinion on multi-parameter strategies for PPPM in cancer is translated into real research and development of PPPM or precision medicine (PM) in cancer.

Non-functional pituitary adenoma (NFPA) seriously affects hypothanamus-pituitary-target organ axis system, with a series of molecule alterations in the multiple levels of genome, transcriptome, proteome, and post-translational modifications, and those molecules mutually interact in a molecular-network system. Meta analysis coupled with IPA pathway-network program was used to comprehensively analyze nine sets of documented NFPA omics data, including NFPA quantitative transcriptomics data [280 differentially expressed genes (DEGs)], NFPA quantitative proteomics data [50 differentially expressed proteins (DEPs)], NFPA mapping protein data (218 proteins), NFPA mapping protein nitration data (9 nitroproteins and 3 non-nitrated proteins), invasive NFPA quantitative transriptomics data (346 DEGs), invasive NFPA quantitative proteomics data (57 DEPs), control mapping protein data (1469 proteins), control mapping protein nitration data (8 nitroproteins), and control mapping phosphorylation data (28 phosphoproteins). A total of 62 molecular-networks with 861 hub-molecules and 519 canonical-pathways including 54 cancer-related canonical pathways were revealed. A total of 42 hub-molecule panels and 9 canonical-pathway panels were identified to significantly associate with tumorigenesis. Four important molecular-network systems, including PI3K/AKT, mTOR, Wnt, and ERK/MAPK pathway-systems, were confirmed in NFPAs by PTMScan experiments with altered expression-patterns and phosphorylations. Nineteen high-frequency hub-molecules were also validated in NFPAs with PTMScan experiment with at least 2.5-fold changes in expression or phosphorylation, including ERK, ERK1/2, Jnk, MAPK, Mek, p38 MAPK, AKT, PI3K complex, p85, PKC, FAK, Rac, Shc, HSP90, NFκB Complex, histone H3, AP1, calmodulin, and PLC. Furthermore, mTOR and Wnt pathway-systems were confirmed in NFPAs by immunoaffinity Western blot analysis, with significantly decreased expression of PRAS40 and increased phosphorylation levels of p-PRAS40 (Thr246) in mTOR pathway in NFPAs compared to controls, and with the decreased protein expressions of GSK-3β and GSK-3β, significantly increased phosphorylation levels of p-GSK3α (Ser21) and p-GSK3β (Ser9), and increased expression level of β-catenin in Wnt pathway in NFPAs compared to controls. Those findings provided a comphrensive and large-scale pathway network data for NFPAs, and offer the scientific evidence for insights into the accurate molecular mechanisms of NFPA and discovery of the effective biomarkers for diagnosis, prognosis, and determination of therapeutic targets.

Data from 758 patients with lung adenocarcinoma were retrospectively collected. All patients had undergone computed tomography imaging and EGFR gene testing. Radiomic features were extracted using the medical imaging tool 3D-Slicer and were combined with the clinical features to build a machine learning prediction model. The high-dimensional feature set was screened for optimal feature subsets using principal component analysis (PCA) and the least absolute shrinkage and selection operator (LASSO). Model prediction of EGFR mutation status in the validation group was evaluated using multiple classifiers. We showed that six clinical features and 622 radiomic features were initially collected. Thirty-one radiomic features with non-zero correlation coefficients were obtained by LASSO regression, and 24 features correlated with label values were obtained by PCA. The shared radiomic features determined by these two methods were selected and combined with the clinical features of the respective patient to form a subset of features related to EGFR mutations. The full dataset was partitioned into training and test sets at a ratio of 7:3 using 10-fold cross-validation. The area under the curve (AUC) of the four classifiers with cross-validations was: (1) K-nearest neighbor (AUCmean = 0.83, Acc = 81%); (2) random forest (AUCmean = 0.91, Acc = 83%); (3) LGBM (AUCmean = 0.94, Acc = 88%); and (4) support vector machine (AUCmean = 0.79, Acc = 83%). In summary, the subset of radiographic and clinical features selected by feature engineering effectively predicted the EGFR mutation status of this NSCLC patient cohort.