An error correction strategy for image reconstruction by DNA sequencing microscopy

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• DNA microscopy offers a powerful approach for spatial omics by inferring molecular geometry directly from sequencing data.
• Experimental artifacts, such as spurious crosslinks and fused nodes, can introduce significant errors in adjacency data, distorting spatial reconstructions.

Purpose of the Study:

• To develop and validate a computational method for correcting common experimental artifacts in DNA microscopy.
• To enhance the accuracy and reliability of spatial reconstructions in spatial omics.

Main Methods:

• The method leverages the principle that spatially proximate molecules should be connected to detect and correct errors.
• It specifically addresses spurious crosslinks (connections between any two nodes) and fused nodes (multiple molecules appearing as one).
• Error correction was tested on simulated data with up to 20% error rates and compared against a read count filter on experimental data.

Main Results:

• The developed method successfully corrected errors in simulated DNA microscopy data, even with substantial error rates.
• It demonstrated superior efficiency in removing experimental artifacts compared to a standard read count filter.
• The correction method effectively identifies and rectifies spurious crosslinks and fused nodes.

Conclusions:

• This computational approach significantly improves the accuracy of spatial reconstructions in DNA microscopy.
• By correcting key experimental artifacts, the method reduces data loss and enhances the overall utility of DNA microscopy for spatial omics.
• Integration of this method is expected to substantially improve spatial omics workflows.