RLMM

RLMM performs genotype calling from Affymetrix SNP array data using a robust linear model applied across multiple chips and multiple SNPs to improve genotype classification accuracy.

Key Features:

• Multi-Chip, Multi-SNP Approach: Processes data across multiple chips and SNPs simultaneously to capture cross-chip and cross-SNP relationships.
• Supervised Learning Algorithm: Trains on large datasets with known genotype labels using supervised learning to calibrate classifiers for new datasets.
• Robust Linear Model: Fits a robust linear model to handle variation in the data for stable genotype classification.
• Mahalanobis Distance for Classification: Uses the Mahalanobis distance to assign genotypes by accounting for the covariance structure among features.
• Normalization for Variance Reduction: Applies normalization to reduce non-biological variance arising from chip-to-chip differences.
• Captures Cross-SNP Similarities: Identifies similarities across genotype groups, probes, and thousands of SNPs to enhance classification accuracy.

Scientific Applications:

• Genotype Calling: Calls genotypes from Affymetrix SNP array data, including Affymetrix 100K SNP arrays, with reported superior performance versus the Affymetrix DM procedure and HapMap public calls.
• Comparative Analysis: Enables comparison of genotype calls against established procedures to assess and improve calling accuracy in genetic studies.

Methodology:

Implements supervised learning on large labeled training sets, fits a robust linear model across multiple chips and SNPs, applies normalization to reduce chip-to-chip variance, and classifies genotypes using Mahalanobis distance while leveraging cross-SNP similarities.