A Symptom-First Graph Framework for a Spectrum-Structured Replacement Psychiatric Taxonomy

This manuscript proposes a symptom-first framework for deriving a replacement psychiatrictaxonomy from the empirical structure of patients’ naturalistically described symptoms. Thepoint of departure is not an inherited diagnostic manual, nor any imposed categorical scheme,but a matrix of symptom co-occurrences obtained from large numbers of subjects. From thatcorrelation structure one constructs a symptom graph: symptoms are vertices, and weightededges encode the degree to which symptoms tend to travel together. A minimum spanning tree(MST) then extracts the simplest connected backbone of the symptom landscape, preserving thestrongest organizing relations while stripping away redundant density. The resulting structureis expected to display neighborhoods, hubs, bridges, bottlenecks, gradients, and spectra ratherthan forced categorical boxes. legacy DSM and legacy ICD categories are therefore not inputs,anchors, or privileged layers in the model. They enter only after the fact as legacy comparators:if parts of the empirical symptom taxonomy reproduce familiar diagnostic distinctions, that isinformative precisely because those distinctions were recovered organically rather than imposed.Once the symptom-derived taxonomy is established, additional layers–course, treatment exposure,treatment response, adverse effects, function, risk, context, and outcome–can be added to producea multilayer clinical framework for prognosis and treatment sequencing. The goal is a disruptivebut clinically intelligible nosology: one that begins with the structure of psychiatric phenomenaas reported and observed, replaces leveling categorical distortions with spectrum-structuredorganization, and then connects that organization to the practical realities of clinical care.