测序平台在母婴肠道菌群结构分析中产生的可能偏倚

Objective: To compare the microbiota analysis results from two next-generation sequencing platforms and their bioinformatics pipelines. Methods: With standard protocols at the same lab, 56 maternal-neonate fecal samples were sequenced and analyzed by using 16S rRNA amplicon sequencing on both Ion Torrent S5-xl and Illumina HiSeq 2500 platforms. The correlation analysis, principal component analysis (PCA), principal coordinate analysis (PCoA) and MRPP analysis were performed to compare the differences in the microbiota structures generated by the two sequencing platforms. Results: The alpha diversity indexes Chao1 (t=l. 96, P = 0. 0011), Observed species (t=2.13, P<0. 0010), PD _whole tree (t=2. 07, P< 0. 0010), Simpson (t=l 87, P=0. 0031) and Good coverage (t=2.32, P<0. 0010) varied significantly between the two platforms except Shannon index (all P<0. 05). At different taxonomy levels, the more species of bacteria sequenced and annotated, the lower the correlation of the relative abundances of bacteria founded between two platforms. The sequencing results were consistent between two platforms. PCA results showed that more than 87% of samples were clustered. According to PCoA, 56 samples of the two platforms were divided into two clusters, and the compliance rate between the two platforms was 71. 43%. The differences between bacterial data generated by the two platforms were tested with MRPP, which showed significant differences at family and genus levels separately (A = 0. 0941, P=0. 0010; A = 0. 0852, P=0. 0021). There was no difference in microbiota composition at family level between two platforms in maternal sample group (A = 0. 0352, P = 0. 0061) but in the neonate group (A = 0. 0063, P = 0. 1491). At genus level, there existed significant differences in microbiota both in maternal and neonatal groups (A = 0. 0216, P = 0. 0042, A = 0. 0981, P = 0. 0010). Conclusion: For the amplicon sequencing of same sample on two different platforms, the relative abundance of microbiota is basically similar, but the diversity and correlation are quite different. To increase reproducibility and reliability in cohort studies, it is important to use the same sequencing platform and corresponding pipeline to reduce systematic errors in microbiome analysis.