Melissa

Melissa implements a Bayesian probabilistic model for jointly clustering and imputing single cell methylomes.

Key Features:

• Joint clustering and imputation: Simultaneously clusters single cell methylomes and imputes missing methylation data in a unified Bayesian framework.
• Local correlation modeling: Accounts for local correlations in methylation patterns via a Generalised Linear Model approach.
• Global similarity modeling: Captures global similarities between cells using a mixture modelling approach.

Scientific Applications:

• Single cell methylome analysis: Enables analysis of single cell DNA methylation data where sparse and noisy measurements are common.
• Cell population identification: Identifies distinct cell subpopulations based on methylation patterns through probabilistic clustering.
• Missing data recovery: Imputes unobserved methylation states to enable downstream analyses that require complete data matrices.

Methodology:

Melissa uses a Bayesian probabilistic model combining a Generalised Linear Model for local correlations with a mixture model for global similarities.