Pro-Maya

Pro-Maya estimates the proportions of constituent cell types in tumor mixtures using partially available DNA methylation data to characterize tumor heterogeneity and support immunotherapy-related analyses.

Key Features:

• Iteratively optimized non-negative matrix factorization framework: Pro-Maya uses an iteratively optimized non-negative matrix factorization (NMF) approach to jointly estimate cell-type proportions and methylation profiles of unknown cell types in blood-derived and solid-tumor samples.
• Handling partial data: The method leverages reference methylation profiles for a subset of known cell types, enabling cellular composition inference without requiring complete reference datasets and mitigating limitations associated with cost and data dropout in experimental profiling.
• Benchmark performance: Across three benchmark datasets, Pro-Maya was reported to outperform five existing methods, accurately inferring proportions for all cell types while recovering methylation profiles for previously unknown components.
• Biological consistency and significance: When applied to four tumor types from The Cancer Genome Atlas (TCGA), Pro-Maya-derived immune cell proportion estimates were broadly consistent with prior studies.
• Facilitating immunotherapy research: By reducing dependence on comprehensive reference methylation panels, Pro-Maya enables systematic exploration of tumor immune composition and supports downstream analyses relevant to cancer immunotherapy.

Scientific Applications:

• Tumor heterogeneity analysis: Deconvolution of mixed-cell tumor samples to quantify cellular composition and characterize heterogeneity.
• Immunotherapy response prediction: Estimation of immune cell proportions within tumors to support association analyses with therapeutic response and outcomes.

Methodology:

Pro-Maya takes DNA methylation profiles from tissue mixtures along with reference methylation profiles for a subset of known cell types and uses an iterative optimization procedure within a non-negative matrix factorization framework to jointly estimate cell-type proportions and methylation profiles for unknown components.