6 Base Calling

The base calling step aims to reduce the number of found puncta to extract the high quality puncta. To do so this step uses a ground truth dictionary. These ground truth dictionaries are stored under the analysis-targeted/ directory and can be chosen with the parameter:

• params.GROUND_TRUTH_DICT

A ground truth dictionary varies depending on the type of cells expected to be seen for a given experiment. The ground truth dictionaries contain two variables

• groundtruth_codes
• gtlabels

The groundtruth_codes are a barcode per cell type and can be used to identify the type of cell in the puncta list generated in the previous puncta extraction stage. The gtlabel has a matrix of string names for each type of cell matching a barcode to name. This stage filters the list of puncta using quantile normalization and checks if a given puncta is in each round or color channel.

This stage produces a basename_results.mat containing 2 variables:

• funnel_numbers
• transcript_objects

The funnel numbers is a 4 element vector consisting, in order, of:

1. the number of puncta found in puncta extraction step
2. the number of puncta after filtering
3. the number of puncta with matching barcodes in the given ground truth dictionary
4. the number of false positives (puncta not in ground truth)

The transcript_objects contains a structure for each puncta with a matching barcode. The structure contains an image transcript, position, the voxel shape, and the name of the type of cell.