A rapid simultaneous quantification method of the screening indicators for thalassemia with near-infrared spectroscopy

Thalassemia is a global popular, serious harm genetic disease. Gene carrier’s rates are as high as 24.50% and 11.07% in Guangxi and Guangdong population located south China, respectively. Mean corpuscular hemoglobin (MCH) and mean corpuscular volume (MCV) are preliminary thalassaemia screening indicators, and hemoglobin A2 (HbA2) was used to further distinguish α- or β-thalassemia. These indicators are usually detected by several instruments with chemical reagents. In this study, a rapid simultaneous quantification method of MCH, MCV and HbA2 with NIR spectroscopy was developed, in which the relative indicator (HbA2) was indirectly quantified by simultaneous analysis of two absolute indicators (Hb and Hb•HbA2). Equidistant combination partial least squares (EC-PLS) was proposed and successfully employed here.
A total of 241 samples of human peripheral blood were collected and measured Hb, MCH, MCV and HbA2 values with BC-3000Plus Blood Cell Analyzer and VARIANTTM Hemoglobin Testing System. They are identified as 93 β-thalassemia (positive) and 148 non-thalassemia (negative) by cut-off values (MCH<27.0pg or MCV<80.0fL and HbA2>3.5%). Each sample was configured to 2× dilute hemolytic solution with distilled water, and used to spectrum measurement with XDS Rapid ContentTM Liquid Grating Spectrometer with 2 mm cuvette and 780-2498 nm region (2 nm interval). First, 96 samples were randomly selected as validation set (58 negative and 38 positive). The remaining 145 samples were used as modeling set (90 negative and 55 positive), and further randomly divided 50 times into calibration (50 negative and 30 positive) and prediction (40 negative and 25 positive) sets.
EC-PLS focused on selection of equidistant wavelengths combination, which was a generalization of moving window PLS (MW-PLS). The parameters included initial wavelength (I), number of wavelengths (N), number of wavelength gaps (G), and number of PLS factors (F), and were set as 780-1882&2078-2348 for I, 1-200 for N, 1-10 for G, and 1-30 for F. PLS model was established for each combination (I, N, G, F) and each division (i) of calibration and prediction sets. Root-mean-square errors and correlation coefficients for prediction were denoted as SEPi and RP,i. The mean value and standard deviation of SEPi and RP,i for all divisions were denoted as SEPAve, RP,Ave, SEPSD, and RP,SD, respectively. SEP+=SEPAve+SEPSD was used as a comprehensive indicator of modeling prediction accuracy and stability.
The parameters were selected by min SEP+. The optimal I, N, G, and F were 856 nm, 16, 1, 16 for Hb, 984 nm, 20, 8, 20 for MCH, 1050 nm, 18, 5, 15 for MCV, and 988 nm, 12, 2, 9 for Hb•HbA2, respectively. For validation set, SEP and RP were 3.50 gL-1 and 0.977 of Hb, 1.22 pg and 0.980 for MCH, 3.53 fL and 0.975 for MCV and 0.38 gL-1 and 0.917 for Hb•HbA2, respectively. Further, SEP and RP for HbA2 were 0.31% and 0.965, respectively. Sensitivity and specificity for β-thalassaemia both reached 100%. The results indicate that NIR prediction was highly accurate for analysis of MCV, MCH and HbA2 and for distinguish β-thalassemia. This method is simple, rapid and suited to large population thalassemia screening.