Fig. 6: Model-based compressed sensing of synaptic connectivity applied to multiple cortical cell-type combinations.

a, Comparison of pyramidal–pyramidal connectivity maps obtained on the same population of neurons using single-target stimulation and model-based compressed sensing (ten-target ensembles, 30-Hz stimulation, connections identified using CAVIaR with NWD). Experiment performed over five planes, image shows z-projection. Agreement between synaptic weights estimated by CAVIaR independently from single-target stimulation and ensemble stimulation trials. b, Map of synaptic connections identified using ensemble stimulation across ten pooled experiments. Dashed cyan circle represents region with 30 μm radius where synaptic connections are most uncertain due to potential photocurrent contamination in opsin-positive postsynaptic neurons (however, see Extended Data Fig. 9 and Supplementary Fig. 9 for the suppressive effect of NWD). Identified connections split by plane shown in Supplementary Fig. 10. c, Number of connections identified by CAVIaR using both single-target and ensemble stimulation across ten experiments. Dashed gray line shows identity. d, Agreement between connectivity maps obtained using single-target stimulation compared to model-based compressed sensing. Mean precision, 0.76; mean recall, 0.77. Gray triangles indicate mean values. e–h, same as a–d, but for mapping pyramidal–PV connections over six experiments. Blue circle represents postsynaptic PV neuron. Mean precision, 0.82; mean recall, 0.67. Lower recall for pyramidal–PV connections could arise due to the faster and more irregular PSC kinetics encountered compared to PV–pyramidal connections. i–l, Same as a–d, but for mapping SST–pyramidal connections over six experiments. Mean precision, 0.87; mean recall 0.78. For the box plots in d,h,l, the bounds of the boxes represent the interquartile range, the triangles show the mean and the whiskers represent the lower and upper limits (excluding outliers).