XenoDisorder: a zero-code computational biology methodology with AI chain of custody for bounded PTM-aware disorder reanalysis

XenoDisorder implements the author's Fractal Custody Object and Fractal Custody Graph (FCO/FCG) design in a computational biology framework developed through artificial-intelligence agents. "Zero-code" describes the operator workflow: the author specifies the biological question, analytical design, evidence boundary, statistical comparison, and release gate without manually implementing the pipeline; AI agents implement and execute the resulting software and analyses. The framework reanalyzes cited, externally generated wet-lab-derived datasets and published computational evidence using standard statistical methods. Its current scientific surface is bounded to sequence- and post-translational-modification-aware disorder scoring, exact modified-row evidence, and local research workflows. Exact-row coverage includes anchored citrullination, phosphoserine, phosphothreonine, and bounded phosphotyrosine contexts, but no PTM currently satisfies both exact-row evidence and validated template-path overlap. A governed CAFA6-derived evaluation is retained as a negative result: the evaluated d512 model achieved AUROC 0.8591 while a one-feature residual baseline achieved 1.0000 on a circular label, for lift -0.14. The reusable contribution from that experiment is the grouped evaluation harness, not predictor utility. The FCO/FCG record makes the declared inputs, implementation, analysis, review, and result disposition replayable and tamper-evident.