Classification and survival prediction for early-stage lung adenocarcinoma and squamous cell carcinoma patients

Abstract

Non-small cell lung cancer (NSCLC) is a leading cause of cancer-associated mortality worldwide. Adenocarcinoma (AC) and squamous cell carcinoma (SCC) are two primary histological subtypes of NSCLC, accounting for ~70% of lung cancer cases. Increasing evidence suggests that AC and SCC differ in the composition of genes and molecular characteristics. Previous research has focused on distinguishing AC from SCC or predicting the NSCLC patient survival rates using gene expression profiles, usually with the aid of a feature selection method. The present study conducted a pre-filtering to identify the genes that have significant expression values and a high connection with other genes in the gene network, and then used the radial coordinate visualization method to identify relevant genes. By applying the proposed procedure to NSCLC data, it was demonstrated that there is a clear segmentation between AC and SCC, however not between patients with a good prognosis and bad prognosis. The focus of discriminating AC and SCC differs from survival prediction and there are almost no overlaps between the two gene signatures. Overall, a supervised learning method is preferred and future studies aiming to identify prognostic gene signatures with an increased prediction efficiency are required.

Keywords: GeneRank; connectivity; non-small cell lung cancer; prognosis; radial coordinate visualization.

Figure 1.

Study schema.

Figure 2.

Graphical illustrations of the best RadViz projections under four scenarios (the first 200/500/1,000 genes, or all genes) for the AC/SCC segmentation. The genes were ordered decreasingly based on their GeneRanks, the first 200/500/1,000-gene scenarios include the highest ranked 200/500/1,000 genes, respectively. The table below those projections gives the resulting gene lists and the predictive statistics using 5-fold cross-validations.

Figure 3.

Graphical illustrations of the best RadViz projections under four scenarios (the first 200/500/1,000 genes, or all genes) for the high risk and the low risk segmentation. The table below those projections gives the predictive statistics of the resulting gene signatures using 5-fold cross-validations.

Figure 4.

Heatmaps of the resulting 8-gene diagnostic and prognostic signatures under the first 1,000-gene scenario: (A) For the diagnostic signature. According to the hierarchical clustering, AC and SCC can be separated using the 8 diagnostic genes. (B) For the prognostic signature. According to the hierarchical clustering, these samples could be stratified into three clusters-patients with high risk of death, patients with low risk, and patients with ambiguous labels.

Figure 5.

Venn-diagram of the 3-gene diagnostic signatures under four scenarios. The venn-diagram illustrates that the stability of those 3-gene diagnostic signatures are also good. The numbers in blankets are the ranks of corresponding genes given by the GeneRank method.