scHiCSRS

scHiCSRS improves single-cell Hi-C (scHi-C) data by imputing missing contacts and distinguishing structural zeros from sampling/dropout zeros to enhance analysis of cell-to-cell variability in chromatin interactions.

Key Features:

• Self-Representation Smoothing Method (scHiC-SRS): Leverages spatial dependencies in 2D scHi-C contact matrices and borrows information from similar single cells to impute missing contacts.
• Gaussian Mixture Model: Models observed zeros as a Gaussian mixture to distinguish structural zeros from sampling/dropout zeros.
• High Sensitivity and Accuracy: Demonstrated high sensitivity for identifying structural zeros and accurate imputation of dropout values in simulation studies and real datasets.

Scientific Applications:

• Improved Clustering: Enables more precise clustering of single cells by improving scHi-C data quality.
• Downstream Analysis Enhancement: Enhances downstream analyses that depend on contact matrix completeness and correct interpretation of zeros.

Methodology:

Combines self-representation smoothing (scHiC-SRS) that accounts for 2D spatial dependencies and borrows information across similar single cells with a Gaussian mixture model to distinguish structural zeros from sampling/dropout zeros.