Phenylacetyl glutamine: a novel biomarker for stroke recurrence warning

Abstract

Abstract Background Stroke is the second leading cause of disease-related death and the third leading cause of disability worldwide. However, how to accurately warn of stroke onset remains extremely challenging. Recently, phenylacetyl glutamine (PAGln) has been implicated in the onset of stroke, but evidences from cohort studies of onset are lacking, especially in patients with first-onset or recurrent. It is necessary to deeply demonstrate the effectiveness of PAGln level on warning stroke onset. Methods One hundred fifteen first onset stroke patients, 33 recurrent stroke patients, and 135 non-stroke controls were included in the analysis. Risk factors associated with stroke attacking were evaluated, and plasma PAGln levels were detected via HPLC-MS based method. LASSO regression, Pearson correlation analysis, and univariate analysis were carried out to demonstrate the associations between PAGln levels and risk factors of stroke. Random forest machine learning algorithm was used to build classification models to achieve the distinction of first-onset stroke patients, recurrent stroke patients, and non-stroke controls, and further demonstrate the contribution of PAGln levels in the distinction of stroke onset. Results The median level of PAGln in the first-onset stroke group, recurrent stroke group, and non-stroke group was 933 ng/mL, 1014 ng/mL, and 556 ng/mL, respectively. No statistical correlation was found between PAGln level and subject’s living habits, eating preferences, and concomitant diseases (hypertension, hyperlipidemia, and diabetes). Stroke severity indicators, mainly age and NIHSS score, were found associate with the PAGln levels. Machine learning classification models confirmed that PAGln levels, as the main contributing variable, could be used to distinguish recurrent stroke patients (but not first-onset stroke patients) from non-stroke controls. Conclusion PAGln may be an effective indicator to monitor the recurrence in stroke patients.

Keywords

• Phenylacetyl glutamine
• Stroke
• LASSO regression
• Risk factors
• Machine learning