methyLImp2

methyLImp2 estimates missing values in DNA methylation data using linear regression-based imputation with chromosome-wise parallelization and mini-batch processing for improved computational efficiency.

Key Features:

• Linear regression imputation: The method leverages high inter-sample correlation of methylation levels to estimate missing values through linear regression.
• Chromosome-wise parallelization: Implementation is parallelized over chromosomes, exploiting the independence of probes on different chromosomes to reduce runtime by several 10-fold.
• Mini-batch approach: A subset of samples is used for imputation in large datasets, reducing runtime from days to hours or minutes without sacrificing prediction performance.

Scientific Applications:

• DNA methylation data imputation: Estimation of missing values in DNA methylation datasets, enabling analysis of incomplete data from methylation arrays or sequencing platforms.
• Large-scale methylation studies: Handling of large datasets with numerous missing values that were previously computationally unfeasible to impute.

Methodology:

methyLImp2 is based on linear regression imputation. The implementation introduces chromosome-wise parallelization and a mini-batch approach that uses only a subset of samples for imputation.