Intensity extraction and crosstalk attenuation using interpolation and adaptation for base calling

A non-transitory computer readable storage medium impressed with computer program instructions to perform base calling, the instructions, when executed on a processor, implement a method comprising:

• accessing a first section from a plurality of sections of a first image output by a flow cell during a first sequencing cycle of a plurality of sequencing cycles of a sequencing run, the first section of the first image including a first plurality of pixels depicting intensity emission values from a plurality of clusters within the flow cell;
• processing the first section of the first image that is output during the first sequencing cycle, to update coefficients of a corresponding convolution kernel in a plurality of convolution kernels,
• wherein the coefficients of the corresponding convolution kernel are updated by evaluating errors between ground truth feature values of one or more clusters and output feature values of the one or more clusters generated by processing the first section of the first image using previous values of the coefficients;
• accessing a second section from a plurality of sections of a second image output by the flow cell during a second sequencing cycle of the plurality of sequencing cycles of the sequencing run, the second section of the second image including a second plurality of pixels depicting intensity emission values from the plurality of clusters;
• convolving the second section of the second image with the updated coefficients of the corresponding convolution kernel, to generate a feature map comprising a plurality of feature values;
• interpolating one or more feature values adjoining a center of a target cluster to generate a weighted feature value for the target cluster;
• and
• processing the weighted feature value assigned to the target cluster, to base call the target cluster.

A system including one or more processors coupled to memory, the memory loaded with computer instructions to perform base calling, the instructions, when executed on the processors, implement actions comprising:

• accessing a particular section from a plurality of sections of an image whose pixels depict intensity emissions for a first color channel and a second color channel from a target cluster and from locations adjacent to the target cluster;
• convolving a first corresponding convolution kernel in a plurality of convolution kernels with the intensity emissions of the pixels for the first color channel in the image to produce a first feature map, and convolving a second corresponding convolution kernel in the plurality of convolution kernels with the intensity emissions of the pixels for the second color channel in the image to produce a second feature map;
• interpolating (i) one or more feature values of the first feature map adjoining a center of the target cluster, to generate a first weighted feature value for the target cluster, and (ii) one or more feature values of the second feature map adjoining the center of the target cluster, to generate a second weighted feature value for the target cluster;
• and
• processing the first and second weighted feature values assigned to the target cluster, to base call the target cluster;
• wherein the first and second corresponding convolution kernels are trained by applying a cost function and evaluating errors in interpolated weighted feature values, then updating coefficients in the first and second corresponding convolution kernels.