Graph-based Spot Decoding

[gapartel]

Spot decoding strategies based only on intensity look-up of the max-valued channel in each sequencing round can be prone to false decoding when fluorescent spots are highly clustered together or presence of highly structured background fluorescence. One way to solve this issue is to detect spots in all channels and rounds and then decode them jointly in a graph minimizing a given cost function. Depending on what the cost function is based on, it is possible to decode sequences based only on the spatial distance of detected spots or also taking into account the quality of each detection.

I will be willing to implement a decoding algorithm based on a graph model in StarFISH. The class will have the same specifications of the PerRoundMaxChannel class. In order to exploit the graph-model with its full possibilities it would be nice to have a spot detection algorithm with a SpotAttributes that describes the quality of each detection.