Detection of maternal carriers of common α-thalassemia deletions from cell-free DNA

Abstract

α-Thalassemia is a common inherited blood disorder manifested mainly by the deletions of α-globin genes. In geographical areas with high carrier frequencies, screening of α-thalassemia carrier state is therefore of vital importance. This study presents a novel method for identifying female carriers of common α-thalassemia deletions using samples routinely taken for non-invasive prenatal tests for screening of fetal chromosomal aneuploidies. A total of 68,885 Vietnamese pregnant women were recruited and α-thalassemia statuses were determined by gap-PCR, revealing 5344 women (7.76%) carried deletions including αα/--^SEA (4.066%), αα/-α^3.7 (2.934%), αα/-α^4.2 (0.656%), and rare genotypes (0.102%). A two-stage model was built to predict these α-thalassemia deletions from targeted sequencing of the HBA gene cluster on maternal cfDNA. Our method achieved F1-scores of 97.14-99.55% for detecting the three common genotypes and 94.74% for detecting rare genotypes (-α^3.7/-α^4.2, αα/--^THAI, -α^3.7/--^SEA, -α^4.2/--^SEA). Additionally, the positive predictive values were 100.00% for αα/αα, 99.29% for αα/--^SEA, 94.87% for αα/-α^3.7, and 96.51% for αα/-α^4.2; and the negative predictive values were 97.63%, 99.99%, 99.99%, and 100.00%, respectively. As NIPT is increasingly adopted for pregnant women, utilizing cfDNA from NIPT to detect maternal carriers of common α-thalassemia deletions will be cost-effective and expand the benefits of NIPT.

Figure 1

Flow chart of the framework and outcomes.

Figure 2

The human α-globin gene cluster and the mutations detected. The Genes track panel illustrates the α-globin locus on the tip of chromosome 16. The approximate ranges of the four deletions (red bar) detected in this study population (−−^SEA, −α^3.7,  α^4.2 and −− ^THAI) are shown in the “Deletions” track. The ranges of 66 bins (blue bars) within the α-globin gene cluster are shown in the “66 bins” track. The counts of fragments mapped to each bin were used as the raw data for our prediction model. −−^SEA, Southeast Asian deletion; −−^THAI, Thailand deletion; AF, Allele frequency. The figure was drawn using the Gviz package.

Figure 3

Principal component analysis of 6417 samples and their genotypes. Normalized count data of 66 features (bins) from 6417 samples within the training dataset were used for principal component analysis. Only PC1 and PC2 are shown along with the percent of variation explained by each component in brackets. Each dot represents one sample.

Figure 4

Estimated incidence (%) of α-thalassemia phenotypes in the Vietnamese population.