Background: Non-small cell lung cancer (NSCLC) is difficult to treat when metastasis has occurred. This study explores the use of cell-free DNA in the clinical management of NSCLC patients who have Kirsten rat sarcoma viral oncogene homolog (KRAS)-positive mutations and as a marker for prognosis. Methods: Peripheral blood collected from advanced NSCLC patients was examined with digital droplet polymerase chain reaction and ultraviolet spectrometry. KRAS mutations were analyzed and quantitated. The specificity and sensitivity of the proposed assay was computed by associating the results with tumor tissue specimens. Comparison against different sub-groups of patients with different metastatic sites and healthy volunteers were made. Patients were subsequently followed up and survival analysis was conducted. Results: Among the 186 patients recruited, 150 had concordant KRAS mutational profiles using cell-free DNA with tumor tissues. The assay sensitivity and specificity were 80.6% and 100%, respectively. For the 150 patients with concordant results, the range of cell-free DNA quantities in peripheral blood was 5.3 to 115 ng. Among the patient groups with different metastatic sites, we observed that patients with bone metastasis had higher concentrations of cellfree DNA. Survival analysis showed that these patients had worse survival outcome. Patients with higher KRAS counts in peripheral blood also had worse outcome. Conclusion: The use of cell-free DNA presents opportunities for risk stratification of patients and possibly aids in the clinical management of the disease. In the current study for NSCLC, patients with bone metastases showed higher cellfree DNA concentrations. Quantitating the concentrations of cell-free DNA presents a noninvasive biomarker capable of prognostic utility.

No abstract

Domain adaptation aims at extracting knowledge from an auxiliary source domain to assist the learning task in a target domain. When the data distribution of the target domain is different from that of the source domain, the direct use of source data for building a classifier for the target learning task cannot achieve promising performance. In this work, we propose a novel unsupervised domain adaptation method called Feature Selection for Domain Adaptation (FSDA), in which we aim to select a set of informative features. The benefits are two-fold. The first is to reduce the mismatch between the data distributions in the source and target domains by selecting a set of informative features in which they share similar properties. The second is to remove noisy features in the source domain such that the learning performance can be enhanced. We formulate a new sparse learning model for structured multiple outputs, including a vector to select informative features that can be used to jointly minimize the domain discrepancy and eliminate noisy features, and a classifier to perform prediction on the selected features. We develop a cutting-plane algorithm to solve the resulting optimization problem. Extensive experiments on real-world data sets are tested to demonstrate the effectiveness of the proposed method compared with the other existing methods.INDEX TERMS Domain adaptation, feature selection, structured multi-output learning, transfer learning.

No abstract

Feature selection aims to remove irrelevant and redundant features from input data. For gene expression, selecting important genes from gene expression data is essential since the gene expression data often consists of a large number of genes. However, the commonly-used feature selection methods are usually biased toward the highest rank features, and the correlation of these selected features may be high. To overcome these problems, we propose an informative feature clustering and selection method to select informative and diverse genes from the gene expression data. The method consists of two steps. In the first step, a feature clustering (FC) method is designed to cluster total genes into several gene clusters. In FC, a set of feature weights are computed to respect the importance of each gene, and we sort the genes in different gene clusters based on the feature weights. In the second step, we propose a stratified feature selection (SFS) method to select genes from different gene clusters and combine them to form the final feature set. Experiments on several gene expression data demonstrate the superiority of the proposed method over six popular feature selection methods.