Dear computational neuroscience colleagues, I would welcome methodological feedback on a new discussion manuscript: A framework for validating neural signals before aggressive behavior in animal models relevant to autism and schizophrenia https://github.com/netanelcyber/psi-qeeg-bis/blob/6b79e9c39cd0532202659acbbe... The central question is whether activity preceding an observed attack in an awake animal contains information that can be recovered from simultaneous external EEG and used to forecast future attacks beyond behavior, EMG and movement measurements. The proposal considers controls, ASD-relevant, SCZ-relevant and combined experimental conditions. It does not claim that animals have the corresponding human diagnoses; the combined condition requires its own construct validation. The analysis separates baseline differences, ongoing-action correlates and genuine pre-event prediction. Feedback would be particularly valuable on depth-to-surface signal mapping, latent-state models, causal preprocessing, animal-level holdout, calibrated warning probabilities, and false alarms per hour during continuous recording. Relevant existing synchronized depth/EEG/video datasets would also be helpful. This AI-assisted manuscript is a hypothesis and proposed validation framework. It contains no new experimental data, trained predictor or validation results, and has not been peer reviewed or accepted by a preprint server. AI assistance and pending author disclosure fields are stated in the draft. Please send comments privately to nsh531@gmail.com. Best regards, Netanel Stern Independent researcher, Israel