Source-linked AI summary
Reconstruction and Analysis of Cancer-specific Gene Regulatory Networks from Gene Expression Profiles
Khalid Raza, Rajni Jaiswal
TL;DR
The paper addresses reconstructing disease-specific gene regulatory networks from high-throughput gene-expression profiles for cancer analysis. It applies two-stage gene filtering and Pearson correlation to prostate-cancer data, validates the results against databases and literature, and obtains a 29-gene network with 55 regulatory relations. The authors state that further experimental validation is needed.
Problem
The study addresses the need to reconstruct cancer-specific gene regulatory networks from whole-genome expression profiles for disease identification and cancer understanding.
Method
The approach filters significant genes with t-test and fold change, computes Pearson correlations between gene pairs, and validates the resulting relationships biologically.
Results
29 genes and 55 regulatory relations were identified after retaining correlations with absolute value >=0.85.
Takeaways & Limitations
The reconstructed network includes hub genes and validated gene relationships relevant to prostate cancer.
Takeaways & Limitations
The study's utility and reliability require further experimental validation.
Abstract
from arXiv · showhide
The main goal of Systems Biology research is to reconstruct biological networks for its topological analysis so that reconstructed networks can be used for the identification of various kinds of disease. The availability of high-throughput data generated by microarray experiments fueled researchers to use whole-genome gene expression profiles to understand cancer and to reconstruct key cancer-specific gene regulatory network. Now, the researchers are taking a keen interest in the development of algorithm for the reconstruction of gene regulatory network from whole genome expression profiles. In this study, a cancer-specific gene regulatory network (prostate cancer) has been constructed using a simple and novel statistics based approach. First, significant genes differentially expressing them self in the disease condition has been identified using a two-stage filtering approach t-test and fold-change measure. Next, regulatory relationships between the identified genes has been computed using Pearson correlation coefficient. The obtained results has been validated with the available databases and literature. We obtained a cancer-specific regulatory network of 29 genes with a total of 55 regulatory relations in which some of the genes has been identified as hub genes that can act as drug target for the cancer diagnosis.
1. INTRODUCTION
Microarray profiles measure thousands of genes simultaneously, supporting disease analysis despite having many gene variables and few experiments. The study focuses on reconstructing and analyzing a prostate-cancer gene regulatory network.
- Microarray technology measures expressions of thousands of genes simultaneously for studying diseases including cancer.
- Gene expression data support cancer prediction and diagnosis but contain many measured genes across relatively few experiments.
- Gene regulatory networks represent genes as nodes and regulatory interactions as edges in a graph.
- The study constructs and analyzes a prostate-cancer gene regulatory network because reconstructing regulatory mechanisms from expression profiles remains challenging.
2. RELATED WORKS
Prior work uses discrete, probabilistic, and continuous models to represent gene regulatory networks, each with stated limitations. This study emphasizes identifying hub genes and analyzing their topology in prostate cancer.
- Gene regulatory network models help represent gene interactions and environmental or drug effects.
- Discrete Boolean models simplify genes into active or inactive states, causing information loss through discretization.
- Bayesian network models estimate conditional dependencies in directed acyclic graphs but are unsuitable for temporal gene-expression information.
- The study aims to identify highly connected hub genes and analyze the topology of a prostate-cancer network.
3. MATERIALS AND METHODS
The proposed pipeline preprocesses normalized expression data, filters significant genes, correlates gene pairs, removes weak relationships, and validates and analyzes the resulting network.
- The algorithm preprocesses data, identifies significant genes, computes gene-pair relationships, removes weak correlations, visualizes the network, validates it biologically, and analyzes topology.
- Preprocessing handles missing values, duplicate entries, and missing gene names in normalized expression datasets.
- A two-stage filter identifies disease-related genes using an unpaired t-test followed by fold change.
- Fold change compares expression levels between tissue and cultured conditions, with a log2-specific calculation for already transformed data.
- Pearson correlation coefficient estimates linear dependence between gene pairs, with values near -1 or +1 indicating stronger correlation.
- Gene pairs below the selected absolute-correlation threshold are removed to focus on strongly connected genes.
4. RESULTS AND DISCUSSIONS
The prostate-cancer network was built from microarray data by filtering significant genes, retaining strong Pearson correlations, and validating the resulting gene set against biological databases and literature. The analysis identified a 29-gene network with 55 regulatory relations, including four hub genes.
- Network construction: 27575 genes from 181 tissue and 12 cultured samples were screened using a two-stage filtering strategy, beginning with a two-tailed t-test.Genes with p-value<=0.001 were considered significant before further filtering.
- Network construction: 55 strong regulatory relationships among 29 genes remained after Pearson correlation analysis of 101 extracted genes.Absolute correlation values >=0.85 were retained, reducing the data to 0.001% of the original 27575 genes.
- Biological validation: The 29 genes were validated against biological databases and literature, including reports linking BNIP3, ZFAND2B, SRPX2, CSTF1, and KCNE2-related interactions to prostate cancer.The validation table covered individual genes and their families, while the interaction table represented activation and repression relationships.
- Regulatory relations: 52 of the 55 regulatory relations were activators, while 3 were inhibitors.The sign of each correlation was interpreted as activation (+) or repression (−).
- Hub genes: KRT5, BNIP3, GJB5, and KCNE2 were identified as hub genes in the inferred network.KRT5, BNIP3, and GJB5 each had total degree eight, while KCNE2 had total degree six.
5. CONCLUSIONS
The paper presents a statistical approach for reconstructing prostate-cancer gene regulatory networks and identifies correlated gene pairs and hub genes. The findings may support cancer diagnostics and future disease-specific network studies, but require further experimental validation.
- 55 significantly correlated gene pairs involving 29 genes were used to construct a prostate-cancer regulatory network.The approach applied t-tests and fold-change filtering before calculating pairwise correlations and removing weak relationships.
- The constructed network revealed genes functioning as hubs among the cancer-associated regulatory relationships.
- The study reports that its findings can provide insights into cancer diagnostics, prognostics, and therapy.
- The authors state that the study requires further experimental validation because cancer regulatory relationships are difficult to obtain and microarray data are inherently noisy.
- The proposed approach could be applied to construct disease-specific regulatory networks for colon, lung, and breast cancers.