DBNorm: normalizing high-density oligonucleotide microarray data based on distributions.

Area of Science:

• Bioinformatics
• Data Science
• Genomics

Background:

• Rare disease patient data is often fragmented across diverse platforms and time points.
• Aggregating this data necessitates robust normalization for comparability.

Purpose of the Study:

• To introduce DBNorm, an R package algorithm for normalizing heterogeneous patient data.
• To enable accurate comparison of patient records regardless of data origin.

Main Methods:

• DBNorm merges data distributions by fitting functions (Polynomial, Fourier, Gaussian) to individual datasets.
• It uses probabilities from fitted distributions to create a unified global distribution.
• Users can also implement custom fitting functions.

Main Results:

• DBNorm demonstrated superior normalization performance compared to z-score, average difference, quantile normalization, and ComBat.
• Comparative analysis used statistics, visualization, and classification on diverse microarray datasets.
• Experimental results confirmed DBNorm's effectiveness on both self-generated and public data.

Conclusions:

• DBNorm provides better normalization outcomes than existing conventional methods.
• The algorithm shows promise for applications beyond bioinformatics, including broader data integration challenges.