Transactional database transformation and its application in prioritizing human disease genes

Area of Science:

• Bioinformatics
• Computational Biology
• Data Mining

Background:

• Binary matrices, or transactional databases, are used for diverse data, including gene-phenotype relationships.
• Current matrix completion methods struggle to extract meaningful insights from these (0,1)-matrices.
• Understanding the information within "0" and "1" entries is crucial for applications like genomics.

Purpose of the Study:

• To develop a novel algorithm, INDEVI, for independent-evidence-based transactional database transformation.
• To address the limitations of existing methods in inferring significant information from binary matrices.
• To enhance the analysis of gene-phenotype data by uncovering hidden patterns.

Main Methods:

• INDEVI evaluates each entry in a (0,1)-matrix using "independent evidence" derived from maximal supporting patterns.
• The algorithm extracts evidence irrespective of the entry's original value (0 or 1).
• The method focuses on a succinct and effective transformation process.

Main Results:

• INDEVI demonstrated high promise in analyzing a genephenotype database.
• The algorithm effectively ranked candidate genes.
• INDEVI showed potential in predicting unknown disease genes.

Conclusions:

• INDEVI offers a promising approach for transactional database transformation, particularly for biological data.
• The independent-evidence-based method extracts valuable information missed by traditional techniques.
• This algorithm has significant implications for gene discovery and disease gene prediction.