BERNN: Enhancing classification of Liquid Chromatography Mass Spectrometry data with batch effect removal neural networks.

Liquid Chromatography Mass Spectrometry (LC-MS) is a powerful method for profiling complex biological samples. However, batch effects typically arise from differences in sample processing protocols, experimental conditions, and data acquisition techniques, significantly impacting the interpretability of results. Correcting batch effects is crucial for the reproducibility of omics research, but current methods are not optimal for the removal of batch effects without compressing the genuine biological variation under study. We propose a suite of Batch Effect Removal Neural Networks (BERNN) to remove batch effects in large LC-MS experiments, with the goal of maximizing sample classification performance between conditions. More importantly, these models must efficiently generalize in batches not seen during training. A comparison of batch effect correction methods across five diverse datasets demonstrated that BERNN models consistently showed the strongest sample classification performance. However, the model producing the greatest classification improvements did not always perform best in terms of batch effect removal. Finally, we show that the overcorrection of batch effects resulted in the loss of some essential biological variability. These findings highlight the importance of balancing batch effect removal while preserving valuable biological diversity in large-scale LC-MS experiments.

Cerebrospinal fluid biomarker profiling of diverse pathophysiological domains in Alzheimer's disease.

INTRODUCTION: While Alzheimer's disease (AD) is defined by amyloid-ß plaques and tau tangles in the brain, it is evident that many other pathophysiological processes such as inflammation, neurovascular dysfunction, oxidative stress, and metabolic derangements also contribute to the disease process and that varying contributions of these pathways may reflect the heterogeneity of AD. Here, we used a previously validated panel of cerebrospinal fluid (CSF) biomarkers to explore the degree to which different pathophysiological domains are dysregulated in AD and how they relate to each other. METHODS: Twenty-five CSF biomarkers were analyzed in individuals with a clinical diagnosis of AD verified by positive CSF AD biomarkers (AD, n = 54) and cognitively unimpaired controls negative for CSF AD biomarkers (CU-N, n = 26) using commercial single- and multi-plex immunoassays. RESULTS: We noted that while AD was associated with increased levels of only three biomarkers (MMP-10, FABP3, and 8OHdG) on a group level, half of all AD participants had increased levels of biomarkers belonging to at least two pathophysiological domains reflecting the diversity in AD. LASSO modeling showed that a panel of FABP3, 24OHC, MMP-10, MMP-2, and 8OHdG constituted the most relevant and minimally correlated set of variables differentiating AD from CU-N. Interestingly, factor analysis showed that two markers of metabolism and oxidative stress (24OHC and 8OHdG) contributed independent information separate from MMP-10 and FABP3 suggestive of two independent pathophysiological pathways in AD, one reflecting neurodegeneration and vascular pathology, and the other associated with metabolism and oxidative stress. DISCUSSION: Better understanding of the heterogeneity among individuals with AD and the different contributions of pathophysiological processes besides amyloid-ß and tau will be crucial for optimizing personalized treatment strategies. Highlights: A panel of 25 highly validated biomarker assays were measured in CSF.MMP10, FABP3, and 8OHdG were increased in AD in univariate analysis.Many individuals with AD had increased levels of more than one biomarker.Markers of metabolism and oxidative stress contributed to an AD multianalyte profile.Assessing multiple biomarker domains is important to understand disease heterogeneity.

Mass spectrometry in cerebrospinal fluid uncovers association of glycolysis biomarkers with Alzheimer's disease in a large clinical sample.

Alzheimer's disease (AD) is a complex and heterogeneous neurodegenerative disorder with contributions from multiple pathophysiological pathways. One of the long-recognized and important features of AD is disrupted cerebral glucose metabolism, but the underlying molecular basis remains unclear. In this study, unbiased mass spectrometry was used to survey CSF from a large clinical cohort, comparing patients who are either cognitively unimpaired (CU; n = 68), suffering from mild-cognitive impairment or dementia from AD (MCI-AD, n = 95; DEM-AD, n = 72), or other causes (MCI-other, n = 77; DEM-other, n = 23), or Normal Pressure Hydrocephalus (NPH, n = 57). The results revealed changes related to altered glucose metabolism. In particular, two glycolytic enzymes, pyruvate kinase (PKM) and aldolase A (ALDOA), were found to be upregulated in CSF from patients with AD compared to those with other neurological conditions. Increases in full-length PKM and ALDOA levels in CSF were confirmed with immunoblotting. Levels of these enzymes furthermore correlated negatively with CSF glucose in matching CSF samples. PKM levels were also found to be increased in AD in publicly available brain-tissue data. These results indicate that ALDOA and PKM may act as technically-robust potential biomarkers of glucose metabolism dysregulation in AD.

Mass Spectrometry in Cerebrospinal Fluid Uncovers Association of Glycolysis Biomarkers with Alzheimer's Disease in a Large Clinical Sample.

Background: Alzheimer's disease (AD) is a complex heterogenous neurodegenerative disorder, characterized by multiple pathophysiologies, including disruptions in brain metabolism. Defining markers for patient stratification across these pathophysiologies is an important step towards personalized treatment of AD. Efficient brain glucose metabolism is essential to sustain neuronal activity, but hypometabolism is consistently observed in AD. The molecular changes underlying these observations remain unclear. Recent studies have indicated dysregulation of several glycolysis markers in AD cerebrospinal fluid and tissue. Methods: In this study, unbiased mass spectrometry was used to perform a deep proteomic survey of cerebrospinal fluid (CSF) from a large-scale clinically complex cohort to uncover changes related to impaired glucose metabolism. Results: Two glycolytic enzymes, Pyruvate kinase (PKM) and Aldolase A (ALDOA) were found to be specifically upregulated in AD CSF compared to other non-AD groups. Presence of full-length protein of these enzymes in CSF was confirmed through immunoblotting. Levels of tryptic peptides of these enzymes correlated significantly with CSF glucose and CSF lactate in matching CSF samples. Conclusions: The results presented here indicate a general dysregulation of glucose metabolism in the brain in AD. We highlight two markers ALDOA and PKM that may act as potential functionally-relevant biomarkers of glucose metabolism dysregulation in AD.