We employ shotgun proteomics and data-independent acquisition (DIA) mass spectrometry to analyze cerebrospinal fluid longitudinally collected from 14 amyotrophic lateral sclerosis (ALS) patients (8 males and 6 females). We perform three main analyses of these data: (1) examine the intra- and inter-patient protein variability in CSF; (2) explore the association of inflammation with rate of disease progression; and (3) develop a mixed-effects model to best explain the decrease in ALS-Functional Rating Scale (ALS-FRS) score. Overall, the CSF protein abundances are tightly regulated with the intra-individual variability contributing just 4% to the overall variance. In four patients, a moderately significant correlation (p < 0.1) was observed between inflammation and rate of disease progression. Using a least absolute shrinkage and selection operator (LASSO) variable selection, we selected 55 viable peptides for mathematical modeling via a linear mixed-effects regression. We then employed forward selection to generate a final model by minimizing Akaike’s information criterion (AIC). The final model utilized changes in abundance from 28 peptides as fixed effects to model progression of the disease in these patients. These peptides were from proteins involved in stress response and innate immunity.

Graphical abstract

我们采用shot弹枪蛋白质组学和数据独立采集（DIA）质谱分析从14例肌萎缩性侧索硬化症（ALS）患者（男性8例，女性6例）中纵向收集的脑脊液。我们对这些数据进行三个主要分析：（1）检查脑脊液中患者之间和患者之间的蛋白质变异性；（2）探讨炎症与疾病进展速度的关系；（3）建立混合效应模型，以最好地解释ALS-功能评定量表（ALS-FRS）得分的降低。总体而言，CSF蛋白的丰度受到严格的调节，个体内部差异仅占总体差异的4％。在四名患者中，存在中等显着相关性（p 在炎症和疾病进展速度之间观察到<0.1）。使用最小绝对收缩和选择算子（LASSO）变量选择，我们通过线性混合效应回归选择了55种可行的肽进行数学建模。然后，我们通过使Akaike的信息标准（AIC）最小化，采用前向选择生成最终模型。最终模型利用28种肽的丰度变化作为固定效应来模拟这些患者的疾病进展。这些肽来自参与应激反应和先天免疫的蛋白质。

图形概要