Optocoder: computational decoding of spatially indexed bead arrays

Spatial transcriptomics technologies that can quantify gene expression in space are transforming contemporary biology research. Some of such methods use spatially barcoded bead arrays that are optically sequenced by a microscopy setup to detect bead barcodes in space which can be consecutively matched to cell barcodes from the respective single cell sequencing experiment. To have good quality barcodes and a high number of barcode matches in space, robust and efficient computational pipelines are needed to process raw microscopy images and call the bases of bead barcodes accurately. Here, we present Optocoder, a computational pipeline that takes raw optical sequencing microscopy images as input and outputs bead barcodes in space. Optocoder efficiently aligns images, detects beads, and corrects for confounding factors of the fluorescence signal such as crosstalk and phasing before base calling. Furthermore, we implement a machine learning pipeline that is trained using the signal from the beads that match to illumina barcodes in order to predict non-matching bead barcodes which can boost up the number of barcode matches. We benchmark Optocoder using data from an in-house spatial transcriptomics platform as well as data from the Slide-seq method and we show that it can efficiently process both datasets with minimal modification.

Here, the datasets include the images acquired via a two laser microscopy setup during the optical sequencing process . There are four pucks (P1, P2, P3, P4) and for every puck there are 12 images where every image corresponds to one cycle of optical sequencing. Every image is a 6-channel TIFF image.