Cancer risk prediction models are important in identifying individuals at high risk of developing cancer, which could result in targeted screening and interventions to maximize the treatment benefit and minimize the burden of cancer. The cancer-associated genetic variants identified in genome-wide or candidate gene association studies have been shown to collectively enhance cancer risk prediction, improve our understanding of carcinogenesis, and possibly result in the development of targeted treatments for patients. In this article, we review the cancer risk prediction models that have been developed for popular cancers and assess their applicability, strengths, and weaknesses. We also discuss the factors to be considered for future development and improvement of models for cancer risk prediction.

Cardiorenal syndrome (CRS), defined as acute or chronic damage to the heart or kidney triggering impairment of another organ, has a poor prognosis. However, the molecular mechanisms underlying CRS remain largely unknown. The RNA-sequencing data of the left ventricle tissue isolated from the sham-operated and CRS model rats at different time points were downloaded from the Gene Expression Omnibus (GEO) database. Genomic differences, protein-protein interaction networks, and short time-series analyses, revealed fibronectin 1 (FN1) and periostin (POSTN) as hub genes associated with CRS progression. The transcriptome sequencing data of humans obtained from the GEO revealed that FN1 and POSTN were both significantly associated with many different heart and kidney diseases. Peripheral blood samples from 20 control and 20 CRS patients were collected from the local hospital, and the gene expression levels of FN1 and POSTN were detected by real-time quantitative polymerase chain reaction. FN1 (area under the curve [AUC] = 0.807) and POSTN (AUC = 0.767) could distinguish CRS in the local cohort with high efficacy and were positively correlated with renal and heart damage markers, such as left ventricular ejection fraction. To improve the diagnostic ability, diagnosis models comprising FN1 and POSTN were constructed by logistic regression (F-Score = 0.718), classification tree (F-Score = 0.812), and random forest (F-Score = 1.000). Overall, the transcriptome data of CRS rat models were systematically analyzed, revealing that FN1 and POSTN were hub genes, which were validated in different public datasets and the local cohort.