Limitations of Clustering with PCA and Correlated Noise.

It is now common to have a modest to large number of features on individuals with complex diseases. Unsupervised analyses, such as clustering with and without preprocessing by Principle Component Analysis (PCA), is widely used in practice to uncover subgroups in a sample. However, in many modern studies features are often highly correlated and noisy (e.g. SNP's, -omics, quantitative imaging markers, and electronic health record data). The practical performance of clustering approaches in these settings remains unclear. Through extensive simulations and empirical examples applying Gaussian Mixture Models and related clustering methods, we show these approaches (including variants of kmeans, VarSelLCM, HDClassifier, and Fisher-EM) can have very poor performance in many settings. We also show the poor performance is often driven by either an explicit or implicit assumption by the clustering algorithm that high variance features are relevant while lower variance features are irrelevant, called the variance as relevance assumption. We develop practical pre-processing approaches that improve analysis performance in some cases. This work offers practical guidance on the strengths and limitations of unsupervised clustering approaches in modern data analysis applications.

Synergistic effects of GFAP and Aß42: Implications for white matter integrity and verbal memory across the cognitive spectrum.

Background: Plasma glial fibrillary acidic protein (GFAP), an astrocytic biomarker, has previously been linked with Alzheimer's disease (AD) status, amyloid levels, and memory performance in older adults. The neuroanatomical pathways by which astrogliosis/astrocyte reactivity might impact cognitive outcomes remains unclear. We evaluated whether plasma GFAP and amyloid levels had a synergistic effect on fornix structure, which is critically involved in AD-associated cholinergic pathways. We also examined whether fornix structure mediates associations between GFAP and verbal memory. Methods: In a cohort of both asymptomatic and symptomatic older adults (total n = 99), we assessed plasma GFAP, amyloid-ß42 (Aß42), other AD-related proteins, and vascular markers, and we conducted comprehensive memory testing. Tractography-based methods were used to assess fornix structure with whole brain diffusion metrics to control for diffuse alterations in brain white matter. Results: In individuals in the low plasma amyloid-ß42 (Aß42) group, higher plasma GFAP was associated with lower fractional anisotropy (FA; p = 0.007), higher mean diffusivity (MD; p < 0.001), higher radial diffusivity (RD; p < 0.001), and higher axial diffusivity (DA; p = 0.001) in the left fornix. These associations were independent of APOE gene status, plasma levels of total tau and neurofilament light, plasma vascular biomarkers, and whole brain diffusion metrics. In a sub-analysis of participants in the low plasma Aß42 group (n = 33), fornix structure mediated the association between higher plasma GFAP levels and lower verbal memory performance. Discussion: Higher plasma GFAP was associated with altered fornix microstructure in the setting of greater amyloid deposition. We also expanded on our prior GFAP-verbal memory findings by demonstrating that in the low plasma Aß42 group, left fornix integrity may be a primary white matter conduit for the negative associations between GFAP and verbal memory performance. Overall, these findings suggest that astrogliosis/astrocyte reactivity may play an early, pivotal role in AD pathogenesis, and further demonstrate that high GFAP and low Aß42 in plasma may reflect a particularly detrimental synergistic role in forniceal-memory pathways.

Understanding the Added Value of High-Resolution CT Beyond Chest X-ray in Determining Extent of Physiologic Impairment.

BACKGROUND: Sarcoidosis staging primarily has relied on the Scadding chest radiographic system, although chest CT imaging is finding increased clinical use. RESEARCH QUESTION: Whether standardized chest CT scan assessment provides additional understanding of lung function beyond Scadding stage and demographics is unknown and the focus of this study. STUDY DESIGN AND METHODS: We used National Heart, Lung, and Blood Institute study Genomics Research in Alpha-1 Antitrypsin Deficiency and Sarcoidosis (GRADS) cases of sarcoidosis (n = 351) with Scadding stage and chest CT scans obtained in a standardized manner. One chest radiologist scored all CT scans with a visual scoring system, with a subset read by another chest radiologist. We compared demographic features, Scadding stage and CT scan findings, and the correlation between these measures. Associations between spirometry and diffusing capacity of the lungs for carbon monoxide (Dlco) results and CT scan findings and Scadding stage were determined using regression analysis (n = 318). Agreement between readers was evaluated using Cohen's κ value. RESULTS: CT scan features were inconsistent with Scadding stage in approximately 40% of cases. Most CT scan features assessed on visual scoring were associated negatively with lung function. Associations persisted for FEV1 and Dlco when adjusting for Scadding stage, although some CT scan feature associations with FVC became insignificant. Scadding stage was associated primarily with FEV1, and inclusion of CT scan features reduced significance in association between Scadding stage and lung function. Multivariable regression modeling to identify radiologic measures explaining lung function included Scadding stage for FEV1 and FEV1 to FVC ratio (P < .05) and marginally for Dlco (P < .15). Combinations of CT scan measures accounted for Scadding stage for FVC. Correlations among Scadding stage and CT scan features were noted. Agreement between readers was poor to moderate for presence or absence of CT scan features and poor for degree and location of abnormality. INTERPRETATION: In this study, CT scan features explained additional variability in lung function beyond Scadding stage, with some CT scan features obviating the associations between lung function and Scadding stage. Whether CT scan features, phenotypes, or endotypes could be useful for treating patients with sarcoidosis needs more study.

The textures of sarcoidosis: quantifying lung disease through variograms.

Objective: Sarcoidosis is a granulomatous disease affecting the lungs in over 90% of patients. Qualitative assessment of chest CT by radiologists is standard clinical practice and reliable quantification of disease from CT would support ongoing efforts to identify sarcoidosis phenotypes. Standard imaging feature engineering techniques such as radiomics suffer from extreme sensitivity to image acquisition and processing, potentially impeding generalizability of research to clinical populations. In this work, we instead investigate approaches to engineering variogram-based features with the intent to identify a robust, generalizable pipeline for image quantification in the study of sarcoidosis. Approach: For a cohort of more than 300 individuals with sarcoidosis, we investigated 24 feature engineering pipelines differing by decisions for image registration to a template lung, empirical and model variogram estimation methods, and feature harmonization for CT scanner model, and subsequently 48 sets of phenotypes produced through unsupervised clustering. We then assessed sensitivity of engineered features, phenotypes produced through unsupervised clustering, and sarcoidosis disease signal strength to pipeline. Main results: We found that variogram features had low to mild association with scanner model and associations were reduced by image registration. For each feature type, features were also typically robust to all pipeline decisions except image registration. Strength of disease signal as measured by association with pulmonary function testing and some radiologist visual assessments was strong (optimistic AUC ≈ 0.9, p ≪ 0.0001 in models for architectural distortion, conglomerate mass, fibrotic abnormality, and traction bronchiectasis) and fairly consistent across engineering approaches regardless of registration and harmonization for CT scanner. Significance: Variogram-based features appear to be a suitable approach to image quantification in support of generalizable research in pulmonary sarcoidosis.