2/2 B-SNIP: Algorithmic Diagnostics for Efficient Prescription of Treatments (ADEPT) - Resubmission - 1

仅有临床现象学既不能(I)捕捉基于生物学的疾病实体，也不能(Ii)允许 基于神经生物学的个体化治疗处方。B-SNIP财团展示并复制 精神分裂症、分裂情感障碍和伴有精神病的双相情感障碍缺乏神经生物学特征。B- SNIP转变为基于生物标记物同源性对精神病病例进行分组。我们制作和复制 生物同源的精神病生物型(BT1，BT2，BT3)，可能有助于针对精神病的治疗。 这个为期12个月的项目将开发一种时间和资源高效的算法来推导B-SNIP生物型， 即使在资源不足的环境中也可以实施。就像在实验室医学中一样，程序(熟练) 将逐步(临床评估，然后是认知，然后是电生理学)来产生特定的生物型 治疗既可以实施(已建立的干预措施)，也可以评估(新的治疗开发)。 目标1：B-SNIP生物型目前需要专门的实验室检测设备，并进行多种检测 跨多个分数的统计集成。相反，我们将确定产生以下结果的最佳单项措施 最有效和最高概率的生物型成员资格。ADEPT将在(临床、 认知、电生理)和跨域(临床特征为选择哪些认知测试提供信息 通知电生理测试的选择)。在每个阶段，ADEPT将产生一个生物型分类和 自信。这将允许在一定比例的情况下进行生物型测定，即使在实验室检测时也是如此 资源是有限的。目的2：医学评估中的首次接触涉及临床特征。临床 仅特征本身就会产生足以在很小但很重要的子集中进行靶向治疗的生物型区分 患者中(占15%，多数为BT3型)。目标3：认知测试是技术要求最低的实验室 评估，并且是生物类型的强大的鉴别器。B-SNIP使用BACS、停止信号(SST)和 抗眼跳以评估认知能力。加上对临床特征的认知，识别的准确率将达到80% BT3和占所有病例的40%(大部分是BT2，尽管没有BT1和BT2很难区分 电生理学)。患者将接受基于自适应算法的不同认知测试(例如，SST可以 在某些情况下，更适合于生物型测定)。自适应方法保持了分类精度 同时减轻临床医生和患者的负担。目的4：最重要的生物类型区分电生理学 特征是神经对显著刺激的低反应(BT1)和旺盛的非特异性神经活动(BT2)。我们 使用了多种复杂的电生理措施，但我们将确定产生最大收益的测试和措施 高效的生物型分化。将电生理学添加到临床和认知信息中将产生90- 识别所有病例的生物型的准确率为95%。同样，对于给定的患者，我们将自适应地选择特定的 最大限度提高患者分类准确性的电生理措施(例如，P300可能更好 在某些情况下用于生物型测定)。

Clinical phenomenology alone neither (i) captures biologically based disease entities, nor (ii) allows for individualized treatment prescriptions based on neurobiology. The B-SNIP consortium showed and replicated that schizophrenia, schizoaffective, and bipolar disorder with psychosis lack neurobiological distinctiveness. B- SNIP transitioned to subgrouping psychosis cases based on biomarker homology. We produced and replicated biologically homologous psychosis Biotypes (BT1, BT2, BT3) that may assist treatment targeting for psychosis. This twelve-month project will develop a time and resource efficient algorithm for deriving B-SNIP Biotypes that can be implemented in even under-resourced environments. Like in laboratory medicine, the procedure (ADEPT) will be stepwise (clinical evaluation, then cognition, then electrophysiology) to yield Biotypes for which specific treatments can be either implemented (established interventions) or evaluated (novel treatment development). Aim 1: B-SNIP Biotypes currently require specialized equipment for laboratory testing, and multiple tests with statistical integration across multiple scores. Instead, we will determine the best individual measures that yield the most efficient and highest probability Biotype memberships. ADEPT will be adaptive both within (clinical, cognitive, electrophysiological) and across the domains (clinical features inform selection of cognitive tests which inform selection of electrophysiological tests). At each stage, ADEPT will produce a Biotype classification and confidence. This will allow for Biotype determination in a proportion of cases even when laboratory testing resources are limited. Aim 2: The first contact in medical evaluation involves clinical characterization. Clinical features alone will yield Biotype discriminations sufficient for treatment targeting in a small but significant subset of patients (15%, mostly BT3). Aim 3: Cognition tests are the least technically demanding laboratory assessments, and are powerful discriminators of Biotypes. B-SNIP uses BACS, Stop Signal (SST), and antisaccades to assess cognition. Addition of cognition to clinical features will yield 80% accuracy for identifying BT3s and 40% of all cases (mostly BT2, although BT1 and BT2 are difficulty to differentiate without electrophysiology). Patients will receive different cognitive tests based on the adaptive algorithm (e.g., SST may be superior for Biotype determination in some cases). The adaptive approach preserves classification precision while reducing clinician and patient burden. Aim 4: The most important Biotype differentiating electrophysiology features are low neural response to salient stimuli (BT1) and exuberant nonspecific neural activity (BT2). We used multiple complex electrophysiology measures, but we will identify tests and measures that yield the most efficient Biotype differentiation. Addition of electrophysiology to clinical and cognition information will yield 90- 95% accuracy for identifying Biotypes for all cases. Again, for a given patient, we will adaptively select the specific electrophysiological measures to maximize classification accuracy for that patient (e.g., P300 may be superior for Biotype determination in some cases).