MultiMAP

MultiMAP performs non-linear dimensionality reduction and integration of multimodal biological datasets to enable joint analysis of single-cell transcriptomics, chromatin accessibility, methylation, and spatial data.

Key Features:

• Algorithm (MultiMAP): Implements the MultiMAP algorithm for dimensionality reduction and multimodal integration.
• Integration Flexibility: Integrates any number of datasets into a common embedding.
• Feature Utilization: Leverages features that may be present in some datasets but absent in others to inform the integrated embedding.
• Non-Linear Mapping: Supports non-linear transformations for more nuanced alignment of complex biological data.
• Customizable Influence: Allows specification of the influence or weight of each dataset within the integrated model.
• Scalability: Scales to large datasets for analysis of high-throughput single-cell modalities.

Scientific Applications:

• Single-Cell Biology: Applied to integrate single-cell transcriptomics, chromatin accessibility, methylation, and spatial datasets for joint analysis.
• Temporal Trajectory Analysis: Used to integrate cellular data along temporal trajectories (e.g., a thymus dataset) to compare transcription factor expression and binding site accessibility during T cell differentiation.
• Revealing Regulatory Kinetics: Reveals patterns in gene expression versus binding site opening kinetics to investigate regulatory mechanisms in processes such as T cell differentiation.

Methodology:

Uses the MultiMAP algorithm to perform non-linear dimensionality reduction and integration of any number of datasets, leverages dataset-specific features including those absent from some datasets, allows specification of dataset influence weights, and can align datasets along temporal trajectories for quantitative comparison of transcription factor expression and binding site accessibility.