Target class reduced test error by 19.8% for 141 turtle cortical voltage responses when linked cells were excluded from fitting

We tested whether target cell class improves voltage response prediction in a released turtle cortex panel. The data had 77 direct excitatory responses onto excitatory cells and 64 onto inhibitory cells. Each test excluded all cell IDs linked through the full response table. Separate target-class means reduced squared error by 19.8% relative to one common mean. The main target was asinh(amplitude/0.1 mV), which retained 20 nonpositive estimates. The 95% range from 9,999 component resamples was 6.6% to 33.3%. All fitting was repeated in each resample. In 19 sources with both target classes, the mean inhibitory-minus-excitatory target response was +0.81 mV, with range +0.11 to +2.04 mV. Other response scales and removal of the largest response kept positive prediction gains. The strongest 15 responses included 14 onto inhibitory targets. The result supports target-specific voltage calibration in this panel. It does not identify synaptic conductances or prove a change in spike propagation. The study is exploratory. Animal IDs, target membrane properties, and raw trials are absent. Resampling assumes independent cell ID components. The package preserves the small source input, code, provenance, all predictions and resamples, figures, checks, and offline repeat steps.

turtle cortex · structure and function · paired recordings · target cell class · voltage response · excitatory synapses · prediction · cell identity · open data