Integration of genomic data analysis for demonstrating potential targets in the subgroup populations of squamous cell lung cancer patients

Abstract

Squamous cell carcinoma (SCC) is the second most frequent histologic subtype of non-small cell lung cancer (NSCLC), causing approximately 400,000 deaths per year worldwide. Although targeted therapies have improved outcomes in patients with adenocarcinoma, the most common subtype of NSCLC, the genomic alterations in SCC have not been comprehensively characterized and no therapeutic agents have been approved specifically for the patients with SCC. Therefore, development of novel therapeutic approaches is urgently needed. Here, we developed an integrative approach, called DLSA, to integrate genomic, epigenomic and transcriptomic data. DLSA stratified SCC patients into distinct survival subgroups and identified the potential molecular drivers in individual survival subtypes. Three subgroups of SCC patients with diverse molecular and clinical characteristics were unveiled through DLSA. Combined analysis of clinical and molecular data on those subgroups suggested that the molecular features in the stratified subgroups are not only consistent with the previous findings, but also provide a guide to targeted agents that worth to be evaluated in clinical trials for SCC patients with poor survival. In conclusion, DLSA offers the possibility for faster, safer, and cheaper the development of novel anticancer therapeutics in the early-stage clinical trails.

Deep learning of the splicing (epi)genetic code reveals a novel candidate mechanism linking histone modifications to ESC fate decision

“…Our previously proposed optimization method ( 32 ) were applied to iteratively minimize the optimization problem (formula 2) with respect of weight matrix \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$W$\end{document} and diversity control matrix \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\Psi$\end{document} . By fixing \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\Psi$\end{document} , the optimization problem (formula 2) becomes an unconstrained optimization problem with respect of both vector and sparse matrix.…”

“…Machine learning algorithms, such as k -means clustering, support vector machines (SVM), and random forests (RF) have been applied to these omics data, separately ( 30 ). Recently, deep learning has gained attentions due to its high performance and generalized characteristics for analyzing complex data of various contexts ( 31 , 32 ), such as image and speech recognition ( 33 – 35 ), natural language understanding (arXiv preprint arXiv: 1409.0473 and arXiv: 1603.01417 ), and most recently computational biology ( 12 , 36 – 39 ). It aims to replace handcrafted features with efficient algorithms for unsupervised or semi-supervised feature learning and hierarchical feature representation using a cascade of nonlinear processing unit.…”

“…It aims to replace handcrafted features with efficient algorithms for unsupervised or semi-supervised feature learning and hierarchical feature representation using a cascade of nonlinear processing unit. Our previous work ( 32 ) reveals its potential to capture the latent structures of cancer patients and enable the generation of distinct survival subtypes of cancer patients with unique clinical and molecular characteristics based on those latent structures. Thus, deep learning can also be used to efficiently decipher the splicing (epi)genetic codes involving heterogeneous genetic and epigenetic factors.…”

Expression of Bcl-2 is a favorable prognostic biomarker in lung squamous cell carcinoma

“…Over the past decade, high-throughput detection technology has yielded a mass of tumor data ( 24 ), but these datasets are scattered, due to patient cohort, technology platform and other heterogeneous variables, thus making it hard to compare ( 22 ). In the present study, in order to solve this problem and to make better use of these public database ( 25 ), 901 LUSC gene expression profiles were integrated and the association between expression level and overall survival was analyzed. Furthermore, the prognostic value of Bcl-2 in LUSC was validated with a tissue microarray (TMA) using immunohistochemistry (IHC) analysis.…”

Deep learning of the splicing (epi)genetic code reveals a novel candidate mechanism linking histone modifications to ESC fate decision

“…Machine learning algorithms, such as k-means clustering, support vector machines (SVM), and random forests (RF) have been applied to these omics data, separately (30). Recently, deep learning has gained attentions due to its high performance and generalized characteristics for analyzing complex data of various contexts (31,32), such as image and speech recognition (33)(34)(35), natural language understanding (arXiv preprint arXiv:1409.0473 and arXiv:1603.01417), and most recently computational biology (12,(36)(37)(38)(39). It aims to replace handcrafted features with efficient algorithms for unsupervised or semi-supervised feature learning and hierarchical feature representation using a cascade of nonlinear processing unit.…”