Muscle pairs with shared attachment partners had a 0.012 input cosine excess under shuffles within muscle type, region and side in one Platynereis graph

Do muscles with a common physical attachment partner have similar neural inputs beyond their cell labels? We joined two published maps from the same three-day-old Platynereis dumerilii larva. The fixed chemical export gave 2,353 input synapses from 205 identified neuron skeletons to 288 muscles of 37 types. We tested 5,737 pairs of different muscle types in the same body region and on the same side. Mean input cosine was 0.1484 for 456 pairs with a common non-muscle attachment partner and 0.0562 for 5,281 other pairs. The gap was 0.0922. We shuffled whole input profiles within muscle type, region and side. The exact expected gap was 0.0803. The residual was 0.0119; a one-sided reference test with 9,999 shuffles gave p = 0.0174. Adding exact input partner count to the shuffle blocks left a residual of 0.0024 and p = 0.226. That control could change only 109 of the 456 linked-pair scores, compared with 231 under the type control. It therefore does not establish that partner count explains the gap. Neural edge cutoffs of three and five also gave no clear residual under the type control. The two maps come from one animal. Muscle types were partly defined from neural and attachment links. Input cosine does not measure coactivation, and a shared attachment does not measure force transfer. The contribution is a conditional test, a complete pair atlas, an audit of source export differences, and code with repeat instructions.

Platynereis · neuromuscular connectivity · desmosomes · muscle attachment · conditional permutation · connectome · input similarity