Periodic Reporting for period 2 - DNA_MICROSCOPY (In situ DNA sequencing-based microscopy for subcellular spatial transcriptomics)
Reporting period: 2022-11-01 to 2024-04-30
This project addresses the need for new methods in spatial transcriptomics, the field of molecular biology related to the mapping of genes to their spatial locations in tissues. We aim to develop a new class of spatial transcriptomics methods based on the principle of spatial DNA networks, whose structure can be reconstructed from high throughput sequencing methods, and then related to spatial locations without relying on classical optics or printing methods. This class of technologies would open up new possibilities in molecular imaging that were previously limited by the constraints of optical techniques such as the limited range of fluorescent colors that can be used to distinguish multiple molecular targets. The overall objectives are to develop a prototype chemistry for constructing molecular networks for 2d transcriptomic imaging, a prototype chemistry for 3D molecular imaging, and computational and theoretical methods for reconstructing spatial molecular positions from networks.
So far in the project we have developed an apparatus for producing molecular networks that we can reconstruct computationally from sequencing data according to the principles of imaging-by-sequence or DNA sequencing-based microscopy. We have also performed work on theoretical principles of reconstructing spatial positions from networks, resulting in a breakthrough in our ability to deduce network structure using random walks. This new algorithm is called Spatio-topological recovery by network discovery (STRND). In the field of sequencing based microscopy or imaging-by-sequencing, molecular networks form as a result of chemistry performed to create linkages between randomly barcoded DNA strands. The particular chemistry determines the “rules” by which network structure forms, for example with nearby strands forming linkages vs strands separated by intermediate distances. This influences the subsequent rules for recovering positions computationally. The idea behind STRND is that one can still reconstruct the spatial locations of molecules comprising a molecular network without explicit knowledge of the rule behind the network’s formation. STRND uses random walks and a neural network to recover spatial information from diverse network structures, bypassing the need for explicit knowledge of the network-forming rules. The STRND result has been published in the journal Nanoscale this year.
Our lab has pushed the field of spatial transcriptomics further through our development of optics free molecular localization techniques. In particular, our new reconstruction algorithm STRND represents the field’s most robust and best characterized optics-free imaging-by-sequencing reconstruction algorithm so far, as we have placed an emphasis on comparison with existing algorithms, the development of quality metrics which are new to the field of imaging-by-sequencing. Our aim, going forward, is to apply these computational and theoretical advancements to the physical data that we have been generating using our DNA network-forming chemistry, apply the technique to biological tissues, and to use this methodology for the study of new biological phenomena such as the maternal-to-zygotic transition in development.