Missing value imputation for gene expression data by tailored nearest neighbors.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-dimensional biological data, such as gene expression and RNA-sequencing, frequently contain missing values.
• Missing data can significantly compromise the accuracy of downstream analyses and results.
• Existing imputation methods struggle with the high dimensionality inherent in genomic datasets.

Purpose of the Study:

• To propose a novel imputation procedure for missing values in high-dimensional gene expression data.
• To address the challenges posed by the 'curse of dimensionality' in imputation methods.
• To compare the proposed method against established imputation techniques.

Main Methods:

• Developed a weighted nearest neighbors (WNN) imputation algorithm.
• Calculated distances using only relevant genes to improve imputation accuracy.
• Avoided the 'curse of dimensionality' by focusing on informative gene subsets.

Main Results:

• The WNN method demonstrated superior performance in imputing missing values in high-dimensional data.
• Outperformed mean imputation, KNNimpute, and random forest imputation in simulations and real-world cancer datasets.
• Successfully handled datasets where the number of predictors exceeds the number of samples.

Conclusions:

• The proposed weighted nearest neighbors approach is a robust and effective solution for missing value imputation in high-dimensional genomics.
• This method offers improved accuracy and reliability for analyzing incomplete gene expression and RNA-sequence data.
• The WNN algorithm provides a valuable tool for researchers working with complex biological datasets.