Feature selection is an important task in machine learning, which can effectively reduce the dataset dimensionality by removing irrelevant and/or redundant features. Although a large body of research deals with feature selection in single-label data, in which measures have been proposed to filter out irrelevant features, this is not the case for multi-label data. This work proposes multi-label feature selection methods which use the filter approach. To this end, two standard multi-label feature selection approaches, which transform the multi-label data into single-label data, are used. Besides these two problem transformation approaches, we use ReliefF and Information Gain to measure the goodness of features. This gives rise to four multi-label feature selection methods. A thorough experimental evaluation of these methods was carried out on 10 benchmark datasets. Results show that ReliefF is able to select fewer features without diminishing the quality of the classifiers constructed using the features selected

The feature selection process aims to select a subset of relevant features to be used in model construction, reducing data dimensionality by removing irrelevant and redundant features. Although effective feature selection methods to support single-label learning are abound, this is not the case for multi-label learning. Furthermore, most of the multi-label feature selection methods proposed initially transform the multi-label data to single-label in which a traditional feature selection method is then applied. However, the application of single-label feature selection methods after transforming the data can hinder exploring label dependence, an important issue in multi-label learning. This work proposes a new multi-label feature selection algorithm, RF-ML, by extending the single-label feature selection ReliefF algorithm. RF-ML, unlike strictly univariate measures for feature ranking, takes into account the effect of interacting attributes to directly deal with multi-label data without any data transformation. Using synthetic datasets, the proposed algorithm is experimentally compared to the ReliefF algorithm in which the multi-label data has been previously transformed to single-label data using two well-known data transformation approaches. Results show that the proposed algorithm stands out by ranking the relevant features as the best ones more often.

Paxillin (PXN) gene mutations are associated with lung adenocarcinoma progression and PXN is known to be a target gene of microRNA-218 (miR-218). On this basis, we hypothesized that PXN overexpression via miR-218 suppression may promote tumor progression and metastasis and that PXN may predict survival and relapse in non–small cell lung cancer (NSCLC). Expression of miR-218 and PXN in 124 surgically resected lung tumors were evaluated by real-time PCR and immunohistochemical analysis. The prognostic value of miR-218 and PXN expression on overall survival (OS) and relapse-free survival (RFS) was analyzed by the Kaplan–Meier test and Cox regression analysis. miR-218 expression in lung tumors was negatively associated with PXN expression. Multivariate analyses showed that PXN and miR-218 might independently predict OS and RFS, respectively, in NSCLC. Moreover, patients with low miR-218 combined with PXN-positive had the worst OS and RFS among the 4 combinations. In a cell model, PXN was negatively regulated by miR-218 and cell proliferation, invasion, and soft agar colony formation were enhanced by PXN overexpression induced by miR-218 suppression. Taken together, our findings suggest that PXN overexpression induced by miR-218 suppression is an independent predictor of survival and relapse in NSCLC, highlighting PXN as a potential therapeutic target to improve clinical outcomes in this disease. Cancer Res; 70(24); 10392–401. ©2010 AACR.

Inactivation of p53 by human papillomavirus 16/18 E6 plays a crucial role in cervical tumorigenesis. To investigate the involvement of HPV16/18 in lung tumorigenesis, the association between HPV16 or HPV18 E6 and p53 protein expression in 122 lung tumors was evaluated by immunohistochemistry, and data showed that HPV16/18 E6 expression correlated inversely with p53 expression, which was further confirmed by tissue in situ immunostaining. Real-time reverse transcription-PCR analysis indicated that E6-positive tumors had lower p21^WAF1/CIP1 and mdm2 mRNA levels than E6-negative tumors. To elucidate the role of E6 in p53 inactivation, we successfully established lung adenocarcinoma cell lines with or without HPV16 infection from patients' pleural effusions. Western blotting showed that E6 protein was indeed expressed in HPV16-infected cells and a lower level of p53 protein was observed in E6-positive cells compared with E6-negative cells. Moreover, the levels of p21^WAF1/CIP1 and mdm2 mRNA in E6-positive cells were lower than in E6-negative cells. The interaction of E6 with p53 protein was revealed by immunoprecipitation assay showing that p53 could be inactivated by E6 protein. Conversely, p53 proteins and p21^WAF1/CIP1 and mdm2 mRNA expressions were restored in E6-knockdown cells by RNA interference compared with control cells. These results reveal that HPV16/18 E6 may be partially involved in p53 inactivation to down-regulate p21^WAF1/CIP1 and mdm2 transcription. In conclusion, HPV16/18 E6 is indeed expressed in HPV DNA–positive lung tumors and is involved in p53 inactivation to contributing to HPV-mediated lung tumorigenesis. [Cancer Res 2007;67(22):10686–93]