| Due to the ever-evolving development of microbial detection technology based on high-throughput sequencing(HTS),human microbiome,especially the gut microbiome,has become one of the most studied and interesting topics in the field of basic medical research.This technology also provides a unbiased pathogen detection strategy that does not rely on traditional microbial culture for the diagnosis of clinical infectious diseases.The former identifies and classifies microbes by targeted amplification and sequencing of the hypervariable regions of 16s rRNA genes of bacteria,and is widely used in the study of microbial community profiling in various populations.The latter can sequence the genome DNA of all microbes in a sample at one test,and can not only identify the known species to the species or subspecies level,but also reconstruct the genomes of species that have not been cultured so far by de novo assembly.metagenomic next-generation sequencing(mNGS),which is gradually widely used in the field of clinical infectious disease diagnosis,focuses on the identification of infection-related pathogens in samples.Although the above techniques play important role in microbial detection,they are complex and still lack of standardized detection procedures and available reference materials.A variety of factors in the experimental process affect the stability and accuracy of the test results.In this study,by preparing reference materials that accord with the characteristics of clinical samples and are suitable for high-throughput sequencing techniques,multicenter quality assessment studies were carried out among high-throughput sequencing laboratories in China for microbial community profiling and pathogen identification.The accuracy and the comparability of test results between laboratories were evaluated.The sources of variation affecting the accuracy were analyzed and the possible solutions were discussed.This knowledge is helpful in promoting further optimization,integration and standardization of high-throughput sequencing assays.In the first section,we designed four mock communities and one negative quality routinely performing high-throughput sequencing(including 16Ss and SMs)of gut microbiota.The results showed that there were great differences in the details of the methodology used by each participating laboratory from the sample processing to the final data analysis process.Overall,the microbial composition(relative abundance)in sample 201901 reported by 46.2%(12/26)of 16Ss laboratories and 82.6%(19/23)of SMs laboratories were moderately or highly significantly correlated with the expected results(Spearman r>0.59,p<0.05).The results from laboratories with near-identical protocols showed slight interlaboratory deviations.For every microorganism,the relative abundance observed in different laboratories varied greatly.In the 16Ss laboratory,the range of relative abundance of Bacteroides spp.was from 0.3%to 53.5%,Enterococcus spp.was from 0.8%to 43.9%and Fusobacterium spp.was from 0.1%to 39.8%.In SMs laboratory,The observed relative abundance variation of B.thetaiotaomicron was the largest,ranging from 5.5%to 53%(Fig.ID).SMs was superior to 16Ss in the detection of low abundance bacteria(B.bifidum)in sample 201901 and F.nucleatum associated with colorectal cancer in fecal suspension sample 201904.Principal component analysis(PCA)and PERMANOVA analysis showed that the differences in genomic DNA extraction methods(cell wall-breaking method and DNA extraction kit),targeted amplification regions(suitable for 16Ss)and bioinformatics analysis tools(annotation classification tools and reference databases)were important factors for the inconsistency of results among laboratories.The contamination of exogenous microbes,whether in 16Ss or SMs laboratories,is always an inevitable problem.In this study,it was found that various microbes were detected in negative control samples in 38.5%(10/26)and 30.4%(7/23)laboratories,respectively.The types of contaminated microbes detected in different laboratories were different.The sources of these unexpected microbes might include molecular biological reagents(such as DNA extraction reagents and PCR reagents),researchers'skin,laboratory environments,cross-sample contamination and mismatching of bioinformatics analysis.In the second section,we constructed 11 mock communities(S1-S11)containing 15 respiratory infection-related microbes and evaluated the performance of 90 laboratories that had established mNGS testing workflows.The information from the questionnaires showed that the technical details of the protocol used by each laboratory for testing mock communities were very different.The test results of each laboratory indicated that in the sample S1 with 14 microbes,the proportion of laboratories with positive results for microbes with a concentration of 1×106 or 1×107 cell/ml was more than 92%.By contrast,less than 50%of laboratories could detect microbes with a concentration of 1×103 cell/ml or less.For each microbe,the RPM values reported by each laboratory varied greatly.For example,in sample S1,the median RPM of H.influenzae and S.pneumoniae were 17556.9(range:0.3-330310)and 24977.3(range:0.5-962186.2),respectively.In the detection of low microbial biomass samples(S2-S4),the introduction of host DNA depletion process in mNGS protocols would lead to a decrease in the detection rate of low concentration microbes,while adding a bead-beating process can increase the yield of low concentration microbes,especially for fungi with hard cell walls such as A.fumigatus.When using mNGS to distinguish genetically similar microbes(S.aureus and S.epidermidis)in the sample S5-S7,the calculated RPM ratio(RPM s.aureus/RPM s.epidermidis)in only 56.6%(43/76)to 63.0%(51/81)of the laboratories had a low-level deviation from the theoretical ratio(within the range of a 2-fold change),suggesting that the process of sample handling or data analysis in some laboratories needs to be further optimized for improving the ability to accurately identify and quantify similar species in the samples.In the detection of three simulated case-related samples(S8-S10),only 56.7%(51/90)to 83.3%(75/90)of laboratories provided a clear etiological diagnosis,indicating that some laboratories have poor ability to identify true pathogens from multi-microbial samples combined with the clinical case information and mNGS test results.Like the result found in the first section of this study,the detection of non-target microbes(false positive results)was also an issue in mNGS testing.Up to 306 non-target microbes were reported in 42.2%(38/90)of the laboratories.False positive results may seriously interfere with the identification of true pathogens from sequencing data,especially for low microbial biomass samples,so the laboratories should take measures to control the potential contamination in the experimental process and errors in sequence alignment in bioinformatics analysis.In summary,after fully considering the biological characteristics of clinical samples and the factors affecting high-throughput sequencing,we designed two sets of mock communities as reference materials.With these reference materials,we further carried out two large-scale multicenter studies to evaluate the NGS protocols for microbial community profiling and pathogen identification,respectively.The results showed that there were significant differences among 16Ss and SMs laboratories for microbial community profiling and mNGS laboratories for pathogen identification.By analyzing the reported data of each laboratory,we found the source of variation and discussed the possible solutions,which is helpful to guide the laboratory to perform methodological optimization,integration and standardization,and finally achieve the purpose of improving the laboratory detection ability.The results of this study also revealed that the reference materials were suitable for all kinds of mainstream sequencing platforms and could evaluate the performance characteristics of NGS testing from different aspects.These materials can be used as standards to evaluate the testing ability of different laboratories to ensure the accuracy and comparability of inter-laboratory test results.At the same time,they can be used as quality control samples to help laboratories track the issues existing in the daily testing process and to further promote the optimization and improvement of the testing process. |
PDF Full Text Request