Unicorn: enhancing single-cell Hi-C data with blind super-resolution for 3D genome structure reconstruction

Area of Science:

• Genomics
• Computational Biology
• Structural Biology

Background:

• Single-cell Hi-C (scHi-C) data reveal unique 3D genomic structures but are often sparse and noisy.
• Accurate reconstruction of high-resolution chromosomal structures from scHi-C data is challenging.
• Existing super-resolution methods have limitations in preserving biological patterns and minimizing noise.

Purpose of the Study:

• To introduce ScUnicorn, a novel blind super-resolution framework for enhancing scHi-C data.
• To develop 3DUnicorn, a maximum likelihood algorithm for precise 3D chromosomal structure inference using enhanced scHi-C data.
• To provide a robust computational framework for improving the fidelity of scHi-C data and 3D genome structure reconstruction.

Main Methods:

• ScUnicorn employs an iterative degradation kernel optimization process for super-resolution reconstruction.
• This approach reconstructs high-resolution interaction matrices without relying on downsampling or predefined degradation ratios.
• 3DUnicorn utilizes enhanced scHi-C data to infer 3D chromosomal structures via maximum likelihood estimation.

Main Results:

• ScUnicorn demonstrates superior performance compared to state-of-the-art methods, as evidenced by higher Peak Signal-to-Noise Ratio, Structural Similarity Index Measure, and GenomeDisco scores.
• 3DUnicorn's reconstructed 3D genome structures show strong agreement with experimental 3D-FISH data.
• The combined ScUnicorn and 3DUnicorn framework significantly enhances scHi-C data fidelity and enables accurate 3D genome structure reconstruction.

Conclusions:

• ScUnicorn and 3DUnicorn offer a powerful computational solution for analyzing challenging scHi-C data.
• The framework effectively addresses data sparsity and noise, leading to more reliable 3D genome structure models.
• This advancement facilitates deeper insights into genomic organization and function at the single-cell level.