Source adaptation gave uncertain 50 Hz prediction gains in three selected Allen slice groups

We tested whether source firing adaptation improves prediction of 50 Hz synaptic response change in Allen r2.1 slices. Donor-held-out ridge models controlled cell labels, distance, initial response size, and trial count. Primary mean absolute error gains were -0.00134 in 155 mouse excitatory connections, -0.00053 in 345 mouse inhibitory connections, and 0.01938 in 76 human excitatory connections. All conditional 98.75% donor bootstrap intervals included zero. An alternate donor split gave a positive human excitatory interval, so that result depends on the split. Eight human inhibitory connections were too few to test. Independent scoring and strong synthetic-signal checks passed. Selected slices, missing measurements, different time scales, and limited human donors prevent causal or cross-species conclusions. Original code, aggregate results, hashes, and external retrieval steps support reproduction. External records stay outside the publication.

Allen Institute · firing adaptation · short-term plasticity · donor validation