We then asked if cells maintain a memory of their response levels that changes their responsiveness to future stimulations. Amazingly, the answer is yes for BMP! Progeny of cells with higher BMP response tend to respond more strongly to a second stimulation, up to 3 weeks later.

We then engineered cell lines where the mutation rate in barcode arrays is proportional to WNT or BMP signaling activity. Because we can integrate information from multiple barcode arrays in each cell, we can reliably estimate the history of signaling activity in single cells.

We want to make recorders that are practical and widely accessible. So we developed a ratiometric scheme to drastically simplify in situ barcode readout. We showed that a 2-channel fluorescence microscopy image is enough to infer the number of edits in an array with 12 barcodes.

Molecular recording addresses this issue by linking the activity level of a signaling pathway to mutations/edits in specific sites in the genome, aka barcode arrays. The more signal a cell receives, the more edits accumulate in its barcodes, which can be detected at the endpoint.

Spatial motif analysis can be applied to large datasets of complex tissues, both 2D and 3D. As scalable and multimodal spatial methods continue to develop, we hope that SMORE can help us identify and make sense of intricate spatial organization of cells in complex tissues.

We also noticed that in many cases cells within spatial motifs show significant gene expression differences compared to other cells of the same type, consistent with spatial motifs representing specialized functional units

We then used SMORE to analyze distribution of retinal bipolar cells and found several highly significant spatial motifs. Some expected based on known visual pathways, some revealing new potential mechanisms for how bipolar cells can regulate signal processing in the retina

We tested SMORE’s performance using synthetic graphs with embedded patterns at various frequencies. The good news is we can recover those patterns with high sensitivity and accuracy.

We first developed an algorithm for unbiased sampling of paths from neighborhood graphs. It involves a systematic approach for enumerating all possible paths in a graph. By adding a probability to each step of enumeration, we can uniformly sample all the paths of a given length.

We introduce Spatial Motifs, spatial arrangements of cell types that occur in tissues more frequently than expected by chance, and provide an algorithm for Spatial MOtif REcognition (SMORE).