A survey of neural network-based cancer prediction models from microarray data.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Oncology

Background:

• Microarray data is crucial for understanding gene expression in cancer.
• Neural networks offer advanced capabilities for analyzing complex biological datasets.
• Accurate cancer prediction models are vital for early diagnosis and treatment.

Purpose of the Study:

• To review recent neural network models for cancer prediction using microarray data.
• To highlight the diverse roles of neural networks in cancer analysis.
• To discuss practical considerations for developing neural network-based cancer prediction models.

Main Methods:

• Systematic literature review of articles published between 2013-2018.
• Keyword searches in scientific databases using terms like 'cancer classification' and 'microarray data'.
• Analysis of identified studies focusing on neural network applications in cancer prediction.

Main Results:

• Neural networks are employed for gene expression filtering (data engineering).
• They are used for predicting cancer presence, type, and survival risk.
• Neural networks also facilitate clustering of unlabeled samples.
• Model architecture is dictated by functionality, with hyperparameters often determined through trial-and-error.

Conclusions:

• Neural networks are versatile tools in cancer prediction from gene expression data.
• Their application spans data preprocessing, direct prediction, and sample clustering.
• Effective implementation requires careful consideration of architecture and iterative optimization.