Classifying next-generation sequencing data using a zero-inflated Poisson model.

Area of Science:

• Bioinformatics and Computational Biology
• Genomics and Transcriptomics
• Statistical Genetics

Background:

• RNA-sequencing (RNA-seq) is a high-throughput technique for gene expression analysis, crucial for disease classification.
• Existing statistical methods for microarray data are not directly applicable to discrete RNA-seq data.
• RNA-seq data often exhibit excess zeros, posing challenges for current classification models like Poisson linear discriminant analysis (PLDA).

Purpose of the Study:

• To develop a novel statistical model for accurate classification of RNA-seq data with an excess of zeros.
• To address the limitations of existing methods in handling zero-inflated count data in RNA-seq.
• To improve disease identification and patient stratification using RNA-seq profiles.

Main Methods:

• Proposed Zero-Inflated Poisson Logistic Discriminant Analysis (ZIPLDA) model for RNA-seq data.
• The ZIPLDA model assumes a mixture of a point mass at zero and a Poisson distribution.
• Incorporated a logistic relation for the probability of observing zeros, considering gene means and sequencing depth.

Main Results:

• Simulation studies demonstrated that ZIPLDA outperforms existing methods across various settings.
• Analysis of real RNA-seq datasets (breast cancer and microRNA-seq) confirmed ZIPLDA's superior performance.
• The proposed method effectively handles excess zeros, leading to more accurate classification of RNA-seq data.

Conclusions:

• ZIPLDA is a robust and effective method for classifying RNA-seq data, particularly those with excess zeros.
• The developed method offers an advancement in analyzing RNA-seq data for medical research and disease identification.
• The software for ZIPLDA is publicly available for broader application in the research community.