Diabetic retinopathy （DR） is a common blinding eye disease caused by diabetes mellitus and is the leading cause of acquired vision loss in adults worldwide. DR is asymptomatic in its early stages， and patients often miss the optimal treatment window by the time they seek medical attention due to vision impairment. Traditional methods used for DR diagnosis have inherent limitations and are not conducive to large-scale rapid screening. Biomarkers can reflect the stage of the disease owing to their specificity and sensitivity， which is crucial for the early diagnosis of DR. In the present study， 142 potential literature-based biomarkers associated with DR were included in a knowledge-directed strategy. And metabolites with different physicochemical properties were chromatographically separated using a 100-mm Discovery HS F5 column. A pseudo-targeted metabolomics method based on ultra-high performance liquid chromatography-tandem mass spectrometry （UHPLC-MS/MS） that simultaneously scans positive and negative ions was established to improve analysis coverage and throughput. The method was validated using eight representative isotope-labeled internal standards as analytical targets. All isotope-labeled internal standards exhibited satisfactory linearities in both positive- and negative-ionization modes， with linear dynamic ranges spanning over three orders of magnitude and correlation coefficients （r²） above 0.995. Extraction recoveries ranged between 75% and 108% at three distinct concentration levels with relative standard deviations （RSDs） below 13%. Notably， 91% of the isotope-labeled internal standards exhibited intra-day precision with RSDs of less than 5% across both ionization modes. Similarly， 91% of the analytes demonstrated inter-day precision with RSDs of less than 10%， with all below 16.3%， indicating good method precision. The developed method was used to analyze 137 serum samples， including 40 DR-free patients with diabetes mellitus （NDR） and 97 patients with DR to investigate the practicality of the method. Quality control （QC） samples were used to evaluate the data， which revealed that the instrument was stable during the analytical sequence. Partial least squares discriminant analysis （PLS-DA） models were constructed to identify and differentiate between the metabolic profiles of the DR and NDR groups with the aim of providing a statistical basis for the classification and diagnosis of DR based on metabolic differences observed in the serum samples. The NDR group and DR groups of varying clinical grades were well separated. A total of 85 differential metabolites were identified between NDR and DR groups using nonparametric tests. Further analysis led to the selection of choline and 12-hydroxyeicosatetraenoic acid （12-HETE） as two markers that effectively distinguished the DR and NDR groups. In addition， the markers exhibited a good ability to distinguish between DR and NDR patients when used in combination. The pseudo-targeted metabolomics-based knowledge-directed method developed in this study provides a reference for screening and diagnosing DR.
HOU et al. (2026) studied this question.