Classification of array CGH data using smoothed logistic regression model

Stratifying tumour subtypes based on copy number alteration profiles using next-generation sequence data

MOTIVATION

The role of personalized medicine and target treatment in the clinical management of cancer patients has become increasingly important in recent years. This has made the task of precise histological substratification of cancers crucial. Increasingly, genomic data are being seen as a valuable classifier. Specifically, copy number alteration (CNA) profiles generated by next-generation sequencing (NGS) can become a determinant for tumours subtyping. The principle purpose of this study is to devise a model with good prediction capability for the tumours histological subtypes as a function of both the patients covariates and their genome-wide CNA profiles from NGS data.

RESULTS

We investigate a logistic regression for modelling tumour histological subtypes as a function of the patients' covariates and their CNA profiles, in a mixed model framework. The covariates, such as age and gender, are considered as fixed predictors and the genome-wide CNA profiles are considered as random predictors. We illustrate the application of this model in lung and oral cancer datasets, and the results indicate that the tumour histological subtypes can be modelled with a good fit. Our cross-validation indicates that the logistic regression exhibits the best prediction relative to other classification methods we considered in this study. The model also exhibits the best agreement in the prediction between smooth-segmented and circular binary-segmented CNA profiles.

AVAILABILITY AND IMPLEMENTATION

An R package to run a logistic regression is available in http://www1.maths.leeds.ac.uk/~arief/R/CNALR/.

CONTACT

a.gusnanto@leeds.ac.uk

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

DNA Copy Number Selection Using Robust Structured Sparsity-Inducing Norms

Array comparative genomic hybridization (aCGH) is a newly introduced method for the detection of copy number abnormalities associated with human diseases with special focus on cancer. Specific patterns in DNA copy number variations (CNVs) can be associated with certain disease types and can facilitate prognosis and progress monitoring of the disease. Machine learning techniques have been used to model the problem of tissue typing as a classification problem. Feature selection is an important part of the classification process, because many biological features are not related to the diseases and confuse the classification tasks. Multiple feature selection methods have been proposed in the different domains where classification has been applied. In this work, we will present a new feature selection method based on structured sparsity-inducing norms to identify the informative aCGH biomarkers which can help us classify different disease subtypes. To validate the performance of the proposed method, we experimentally compare it with existing feature selection methods on four publicly available aCGH data sets. In all empirical results, the proposed sparse learning based feature selection method consistently outperforms other related approaches. More important, we carefully investigate the aCGH biomarkers selected by our method, and the biological evidences in literature strongly support our results.

Genetic alterations in oral squamous cell carcinoma progression detected by combining array-based comparative genomic hybridization and multiplex ligation-dependent probe amplification

BACKGROUND

Oral squamous cell carcinoma (OSCC), the most common malignancy of the oral cavity, has been shown to occur via a multistep process driven by the accumulation of carcinogen-induced genetic changes.

STUDY DESIGN

Array-based comparative genomic hybridization (aCGH) and multiplex ligation-dependent probe amplification (MLPA) were conducted to screen human genomewide alterations on fresh tissues of the cancer area, the dysplastic transitional area, and the resection margin (normal) free of tumor; these samples were obtained from 7 OSCC patients.

RESULTS

The highest amplification frequencies (100%, 7/7) were detected in FAM5B, TIPARP, PIK3CA, NLGN1, FGF10, HDAC9, GRM3, DDEF1, EDNRB, CHRDL1, and HTR2C, and the highest deletion frequencies in THRAP3, CTTNBP2NL, GATAD2B, REL, CKAP2L, RHOA, EIF4E3, PDLIM5, FBXO3, NEUROD4, and ABCA5 in the OSCC. In the dysplasia, amplification (100%, 7/7) was detected in RNF36 and deletion in CKAP2L and TCF8. We could detect large differences with MLPA in the number of alterations between the cancer or dysplasia versus the normal area with P values of <.001.

CONCLUSION

These findings indicate that these DNA copy number changes on each chromosome in the 3 categories may be associated with OSCC tumorigenesis and/or progression.