• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    SinoDuplex:An Improved Duplex Sequencing Approach to Detect Low-frequency Variants in Plasma cfDNA Samples

    2020-07-29 05:34:46YongzheRenYangZhangDandanWangFengyingLiuYingFuShaohuaXiangLiSuJianchengLiHengDaiBingdingHuang
    Genomics,Proteomics & Bioinformatics 2020年1期

    Yongzhe Ren ,Yang Zhang ,Dandan Wang ,Fengying Liu ,Ying Fu ,Shaohua Xiang ,Li Su ,Jiancheng Li ,Heng Dai ,Bingding Huang ,,5,*

    1College of Big Data and Internet,Shenzhen Technology University,Shenzhen 518118,China

    2Department of Research and Development,Sinotech Genomics Inc.,Kanxing Road 3399,Shanghai 201314,China

    3Department of Integrated Traditional and Western Medicine In Oncology,The First Affiliated Hospital of Anhui Medical University,Hefei 230022,China

    4Department of Radiation Oncology,Fujian Medical University Cancer Hospital and Fujian Cancer Hospital,Fuzhou 350014,China

    5 Institute of Synthetic Biology,Shenzhen Institute of Advanced Technology,Chinese Academy of Science,Shenzhen 518005,China

    KEYWORDS Next generation sequencing;Liquid biopsy;Circulating tumor DNA;Duplex sequencing;Low frequency variant

    Abstract Accurate detection of low frequency mutations from plasma cell-free DNA in blood using targeted next generation sequencing technology has shown promising benefits in clinical settings.Duplex sequencing technology is the most commonly used approach in liquid biopsies.Unique molecular identifiers are attached to each double-stranded DNA template,followed by production of low-error consensus sequences to detect low frequency variants.However,high sequencing costs have hindered application of this approach in clinical practice.Here,we have developed an improved duplex sequencing approach called SinoDuplex,which utilizes a pool of adapters containing pre-defined barcode sequences to generate far fewer barcode combinations than with random sequences,and implemented a novel computational analysis algorithm to generate duplex consensus sequences more precisely.SinoDuplex increased the output of duplex sequencing technology,making it more cost-effective.We evaluated our approach using reference standard samples and cell-free DNA samples from lung cancer patients.Our results showed that SinoDuplex has high sensitivity and specificity in detecting very low allele frequency mutations.The source code for SinoDuplex is freely available at https://github.com/SinOncology/sinoduplex.

    Introduction

    Liquid biopsies are valuable tools for non-invasive diagnostics and monitoring of diseases such as cancer.These can include sampling of peripheral blood,effusion fluids,and other components of body fluids for extraction of circulating tumor cells(CTCs)[1],tumor-derived cell-free DNA(cfDNA),and other materials(e.g.,exosome,extracellular vesicles)[2].Although PCR-based testing is recommended for routine diagnostics of cancer hotspot mutations [3,4], targeted next-generation sequencing(NGS)using cfDNA,short DNA fragments in the plasma released by apoptotic and necrotic cells, has emerged as the most common approach to assess tumorspecific alterations[5,6].For instance,it has been used to determine the genetic landscape of tumor lesions,monitor treatment responses,track acquired resistance,select existing antiresistance targeting therapies,and assess the presence of residual disease [1,4]. Compared to tissue biopsies, targeted sequencing in liquid biopsies currently appears to be a highly promising and revolutionary tool for diagnosing and monitoring cancer.The method has three major advantages:(i)liquid biopsies enable repeated sampling over a relatively long period of time to monitor the patient’s clinical condition,whereas tissue biopsies cannot be obtained repeatedly or be taken from a specific cancerous lesion location;(ii)liquid biopsies overcome single-biopsy bias,enabling representation of the full extent of tumor heterogeneity;and(iii)PCR-based assays can only detect one or a few known genomic hotspot mutations,which may only apply to a minority of patients.In Asian patients,for example,47%of lung cancer patients had EGFR mutations,compared with only 15%in European patients[7].

    Targeted NGS technology can detect genetic alterations in a wider pool of genomic regions,and has been employed to screen rare mutations in early-stage cancers[8,9],guide targeted therapy,monitor treatment,and enable prognoses[1].However, sub-clonal genetic mutations in tumors may be found in less than 1%of DNA molecules in a plasma sample[10,11].In addition,cfDNA extracted from blood samples usually contain damaged DNA from normal metabolic processes or as a result of DNA extraction[12-14].Consequently,these artefactual mutations might be retained during the PCR amplification process.Additionally,mutations may be misincorporated by DNA polymerase during PCR amplification[15-17]. Therefore, additional refinement for ultrasensitive profiling of circulating tumor DNA(ctDNA)is essential to improve ctDNA quantity and quality,and reduce errors introduced during PCR amplification,library construction and sequencing.To improve sensitivity and minimize errors,different approaches have been developed using unique molecular identifiers(UMIs or molecular barcodes).This allows each DNA molecule to be labeled and accurately tracked.The first barcode strategy was developed to track single DNA strands[18-22].Recently,duplex sequencing(DS),an improved barcoding strategy for tracking double-stranded DNA,was also reported[10,21].DS uses randomly generated barcodes to uniquely tag each DNA fragment in a plasma cfDNA sample.Tagged fragments are then amplified by PCR before being used in the preparation of a sequencing library,creating fragment families characterized by unique combinations of barcodes at both the 5′and 3′ends.A family contains multiple reads,each originating from a single input DNA fragment.A true variant will appear in all reads within a family.In contrast,sequencing and amplification errors will manifest themselves as‘‘polymorphisms”within a family,thus allowing them to be identified and removed by generating consensus sequences.The consensus of all reads originating from the same strand reduces errors originating from PCR amplification and sequencing.Only mutations present in sequences obtained from both complementary DNA strands are counted as true positive mutations. The sequences are referred as duplex consensus sequences(DCS),whereas mutations present in only one of the complementary DNA strands(single-strand consensus sequences,SSCS)are still counted as errors.

    Although DS is expected to drive significant advances and improve sensitivity in detecting rare mutations in ctDNA,the experimental and computational aspects of this technique are still evolving[22-25].In this study,we have developed a novel,efficient DS approach named SinoDuplex,combining a special barcoding strategy consisting of pre-defined barcode sequences with a novel computational algorithm to eliminate background noise and produce more accurate duplex consensus sequencing data.We evaluated the performance of Sino-Duplex with a pool of diluted samples using two reference standard samples,HD701 and HD753,on a targeted panel of 334 genes.SinoDuplex increases the output of DS while reducing cost.At an allele frequency cut-off of 0.1%,SinoDuplex achieved a high sensitivity of 98.62%and specificity of 97.09%.In addition,we applied this method to samples from patients with clinical lung cancer and validated low-frequency hotspot actionable mutations with droplet digital PCR(ddPCR). Our results show that SinoDuplex significantly improves the sensitivity and specificity for the detection of low-frequency variants in plasma ctDNA samples in a costeffective manner.

    Materials and methods

    SinoDuplex adapter synthesis

    Our novel duplex adapters (termed SinoDuplex adapters)employ a pool of at least 16 unique molecular identifiers(UMIs)at the end of the double-strand portion of the Yshaped adapters.The pool of UMIs comprises seven or eight bases of pre-defined and color-balanced sequences mixed in certain ratios to avoid sequencing bias on Illumina sequencing instruments.Every pre-defined UMI differs from all other UMIs by at least three edit distances.SinoDuplex adapters are formed by combining and annealing two single strands of oligonucleotides possessing pre-defined UMI in a pairwise manner.One oligonucleotide is designated as the P5 strand:ACACTCTTTCCCTACACGACGCTCTTCCGATCTXXX XXXX(X)T(where XXXXXXX(X)indicates the position of a fixed 7-or 8-base UMI sequence),and the other is designated as the P7 strand:/5phos/X‘X’X‘X’X‘X’X‘(X’)AGAT CGGAAGAGCACACGTCTGAACTCCAGTCAC (where X‘X’X‘X’X‘X’X‘(X’) indicates the reverse complement of XXXXXXX(X)).(Table S1).Each pair of adapter strands was synthesized(HPLC and NGS grade,Life Technologies,Carlsbad,CA)and combined by equimolar amounts to a final concentration of 100 μM in 1×annealing buffer containing 10 mM Tris,1 mM EDTA and 0.1 M NaCl.Each reaction was heated to 95 ℃for 5 min in an ABI 9700 thermocycler(Applied Biosystems,Foster City,CA)before turning off the machine and leaving the reaction to gradually cool for 1 h.The annealed adapters were finally pooled together in equal volumes and further diluted to 10 μM in Low TE buffer to form a working solution.

    DNA reference sample preparation

    To estimate the performance of our DS strategy,two wellcharacterized genomic DNA reference standard samples,HD701 and HD753 (Horizon Discovery Inc., Cambridge,UK)were used.These samples are commercially available mixtures of DNA from cell lines for which precise allelic frequencies(AFs)of several hotspot actionable mutations have been validated by digital PCR,covering a wide range of mutations including single nucleotide variants(SNVs),indels,fusions,and copy number variation.Allele frequencies of the verified variants in these references were between 1%and 41.5%(details of each variant are provided in Table S2).A dilution series of these two reference samples (HD701M1, HD701M2,HD753M1,and HD753M2)with the HapMap normal cell line NA18536 were generated to simulate different ranges of allele frequencies and assess the performance and limit of detection of our assay.Diluted DNA mixtures were further sheared using a focused-ultrasonicator(S220,Covaris,Woburn,MA)with a target size around 170 bp to mimic the size distribution of cfDNA.

    Patient plasma samples and cfDNA extraction

    A number of liquid biopsy specimens were collected to assist in validation experiments.For each sample,8-10 mL peripheral blood was collected in a cell-free DNA BCT tube(Catalog No. 218962, Streck, La Vista, NE) and centrifuged for 10 min at 1600g at room temperature within three days of drawing blood.The supernatant containing the plasma was further centrifuged at 16,000g for 10 min at room temperature to remove any residual cells.cfDNA was extracted from 4 to 5 mL of plasma and eluted into 50 μL of buffer AVE using the QIAamp Circulating Nucleic Acid Kit (Catalog No.5114,Qiagen,Hilden,Germany)according to manufacturer’s instructions. Quantification of extracted cfDNA was performed using the Qubit 3.0(Thermo Fisher Scientific,Waltham,MA).

    Panel design

    To apply SinoDuplex in a clinical setting,we designed three different targeted gene panels to detect low-allele-frequent mutations in plasma ctDNA samples.The smallest panel is a specially designed panel for lung cancer named LungCore which covers 10 core genes that have targeted drugs approved by the US FDA for lung cancer and actionable fusion events.The second panel,ActionAll,covers 73 genes with actionable mutations in targeted clinical therapy for all solid tumors.The third panel is a pan cancer panel covering the exons of 334 genes and is capable of estimating blood tumor mutational burden (TMB) to select cancer patients suitable for immunotherapy.The genes in this pan-caner panel are selected to cover variants associated with targeted cancer therapies(i)approved by the FDA or listed in the NCCN guidelines,(ii)reported as responsive to therapy in public databases and the literature. Hotspot actionable fusion introns were also included in these three panels to identify actionable fusion events.Capture probes were ordered from Integrated DNA Technologies,Coralville,IA.Gene lists and hotspot introns of these three panel are provided in Table S3.

    Targeted library construction and sequencing

    Pre-capture library preparation was performed using the KAPA Hyper Prep kit(Catalog No.K8504,Roche,Basel,Switzerland)with SinoDuplex adapters.In brief,20-33 ng cfDNA or sheared reference DNA mixture, representing 6000-10,000 haploid genomic equivalents,was used for end repair and A-tailing,followed by ligation of SinoDuplex adapters.Ligated products were bead-purified and further amplified for seven cycles using KAPA HiFi HotStart ReadyMix with unique dual indexes(UDIs),primers that mitigate sample mis-assignment due to index hopping. For each library,500 ng DNA with different UDIs were pooled together.Up to 2 μg of total library was used as input for in-solution capture enrichment with xGen Lockdown Reagents kit(Catalog No.1072281,Integrated DNA Technologies,Coralville,IA)and customized xGen lockdown panels.A hybridization mixture of pooled libraries and customized xGen lockdown probes in xGen Hybridization Buffer was denatured at 95°C for 5 min,then incubated at 65°C for 4-16 h with the addition of human Cot-1 DNA(Catalog No.15279011,Life Technologies,Carlsbad,CA)and xGen Universal Blockers-TS Mix(Catalog No.1075475,Integrated DNA Technologies,Coralville,IA).After incubation,library-probe duplexes were captured with Dynabeads M270 Streptavidin (Catalog No.65306,Invitrogen,Carlsbad,CA)and off-target library fragments were washed off.Bead-captured libraries were further amplified with universal P5/P7 primers in KAPA HiFi HotStart ReadyMix,followed by purification with beads.Postcapture libraries were quantified by Qubit 3.0.Fragment size was determined by a 2100 Bioanalyzer using a High Sensitivity DNA chip(Catalog No.5067-4626,Agilent Technologies,Santa Clara,CA).Paired-End 150 sequencing was performed using the HiSeq X Ten platform(Illumina,San Diego,CA)supporting dual indexing with raw sequencing depths over 20,000×.

    Duplex consensus sequence generation

    The complete computational workflow of the SinoDuplex approach is illustrated in Figure 1.Firstly,the QC and adapter trimming step is performed using fastp[26]to remove adapter contamination and filter out low-quality reads.Simultaneously,UMI barcode sequences are extracted from the read sequences and appended into the read name.Next,read pairs are mapped to the reference genome hg19 using the BWA mem algorithm[27]with‘‘-C”option to set the UMI barcode sequences as a special‘‘BC:Z”tag for each alignment in the bam file.Single-stranded consensus sequences(SSCS)are generated by loading all read pairs with the same mapping positions and orientations into memory and grouping them into different SSCS families with the same UMI barcode sequences at both ends and the same CIGAR string.Unlike the original method,which employs the majority-based algorithm[21],we apply a Bayesian algorithm to determine the final high-quality base at each position of the SSCS consensus sequence and calculate the corresponding consensus quality score for this base using the following Equation(1).

    Figure 1 Schematic illustration of SinoDuplex workflow

    Recall that each base b is associated with a base call and quality score pair(bi,qi)in the consensus group.The posterior probability of the consensus base given the entire consensus group is

    The base with maximum posterior probability is taken as the consensus base and its corresponding quality score is calculated according to the chosen probability.

    After construction of the SSCS sequence and quality score from all SSCS families for each mapping position,our algorithm merges two SSCS read pairs with transposed UMI barcode sequences and the same mapping position into one DCS,if possible.If both bases at the same position in two SSCSs(forward and reverse strands)match,the given base is used and an average Phred quality score is assigned to this base.However,in the case of a mismatch between two bases,the nucleotide‘‘N”is placed in final DCS sequence and a predefined low Phred score(10 is used here)is assigned.If the proportion of N bases in the final DCS sequence is greater than a pre-defined cut-off(50%in this study),this DCS sequence is filtered out.If the corresponding transposed read pair cannot be found for one SSCS,this SSCS read pair is also kept when its family size is greater than 1.The sequence and quality scores of all SSCS and DCS read pairs are written out into two intermediate FASTQ files to release memory. These DCS and SSCS read pairs are then re-mapped onto the human hg19 reference genome using the BWA mem algorithm after processing all raw read pairs using the algorithm above.A final bam file containing both SSCS and DCS read pairs is obtained for further variant calling.

    Variant calling

    For variant calling of ctDNA samples,we used an algorithm based on samtools mpileup[28]developed in-house to detect somatic SNVs and indels[29,30].Our calling algorithm can detect rare somatic mutations with a frequency as low as 0.1%.Briefly,many candidate SNVs/indels were identified in tumor samples with at least three reads and the required mapping quality and base quality score.After querying against a normal reference sample with filtering conditions, eligible mutations were categorized as either germline or somatic.In the calling process,a series of filters were applied on the raw SNV/indel calls,including noise estimation from known SNPs,strand bias filtering, and noise filtering from neighboring regions to ensure reliable variant detection.In our experience,most false positive variants originate from alignment errors and repeat regions.These variants can be removed using a blacklist containing common mistakes from a pool of normal samples. Final high-confidence variants (SNVs and small indels)were then annotated with UCSC RefSeq gene information,dbSNP[31],1K Genome[32],ExAC[33],GnomAD[33],COSMIC[34],and Clinvar[35]using SNPEff[36]and an inhouse database-annotation module based on HTSlib(http://www.htslib.org/)library.

    Performance calculation

    To evaluate the detection capability and limitation of our algorithm,we calculated the sensitivity or positive percentage agreement(PPA)and specificity or positive predictive value(PPV)using the confirmed somatic mutations in reference standards HD701 and HD753 with different dilution ratios using the HapMap normal cell line NA18536. Mutations called in two undiluted samples of HD701 and HD753 with a variant allele frequency >1%were treated as a true mutation data-set and were manually reviewed using the Integrative Genomics Viewer[37].The allele frequencies of some hotspot actionable mutations in these two reference samples were already confirmed by ddPCR assays by Horizon Discovery,Cambridge,UK.Figure S1 shows a high correlation between the AFs detected by SinoDuplex and AFs confirmed by ddPCR for those hotspot actionable mutations in undiluted samples of HD701 and HD753(true mutation data-set).Mutations detected in diluted samples that are also present in the true mutation data-set are treated as true positive(TP);those that are not,as false positive(FP).Mutation in the true mutation data-set that are not detected in the diluted sample are classified as false negative(FN).Thus,sensitivity and specificity are calculated as TP/(TP+FN)and TP/(TP+FP).Detailed information of all mutations(TP,FN,TP)for the diluted samples are provided in Table S4.To determine the appropriate sequencing depth,we performed in silico downsampling of diluted samples and calculated sensitivity and specificity at different depths using the same procedure.

    Droplet digital PCR(ddPCR)validation

    Hotspot actionable mutations such as KRAS(G12D)and EGFR(790M,L858R,and 19Dels)detected in patient cfDNA samples by SinoDuplex were further validated by ddPCR.In brief,two probes targeting mutant and wild-type alleles werelabeled with FAM and VIC dyes,respectively.A TaqMan PCR reaction using ddPCR Supermix for Probes(Catalog No.1863024,Bio-Rad,Hercules,CA)was subjected to droplet generation using the QX200 Droplet Digital PCR system(Bio-Rad,Hercules,CA),followed by PCR.Droplets were analyzed with the QX200 Droplet Reader(Bio-Rad,Hercules,CA)and QuantaSoft software(Bio-Rad,Hercules,CA)for fluorescent measurement and allele calling.

    Table 1 The performance of SinoDuplex at different AF cut-offs:0.1%,0.2%,and 0.5%

    Results

    Improvement of low-frequency variant identification

    To assess the performance of our approach(Figure 1)in detecting low-frequency variants,we calculated the sensitivity and specificity for a pool of diluted samples with known variants in the reference standard samples HD701 and HD753.To simulate a serial dilution of different allele frequency variants,HD701M1 and HD701M2 were diluted 3-fold and 6-fold with the HapMap normal sample NA18536. HD753M1 and HD753M2 were diluted 5-fold and 10-fold with NA18536.All diluted samples were sequenced with the pan-cancer panel,which covers the exons of 334 cancer-related genes and hotspot fusion introns.Performance results for these four diluted samples are summarized in Table 1 with allele frequency(AF)cutoffs of 0.1%,0.2%,and 0.5%.The SinoDuplex approach achieved a high sensitivity of 98.62% and specificity of 97.09%at an AF cut-off of 0.1%.Furthermore,a high correlation between the detected AF and expected AF was observed for all mutations in these four diluted samples(Figure 2).These performance assessment results demonstrated the ability of our approach to detect very low-frequency variants in plasma cfDNA samples.

    Better utilization of raw sequencing data

    In the process of generating SSCS consensus sequences,majority-based rule is often adopted to determine the consensus base at each position.At least three member reads in a SSCS family are needed to form a SSCS sequence in the original DS approach [10,21]. However, this majority-based approach is far from precise and many SSCS families with less three members are technically thrown away.In our Bayesianbased algorithm,we used Equation(2)to calculate the probability of each possible consensus base,selecting the base with maximum probability.Therefore,in our approach,even if there were only two reads in one SSCS family,we were still able to generate high-quality consensus sequences and thus more SSCS reads compared to the original approach.At the same time,we calculated the consensus base quality score using Equation(3).After the consensus sequences were generated,the mean base phred quality score increased from 38 to 80,indicating a significant reduction in the sequencing error rate(Figure S2).

    In addition to reducing the cut-off of the SSCS family size from three to two,several modifications were implemented in the generation of DCS from two complementary SSCS families.If an SSCS family with only one read pair can form a DCS with another SSCS family,it is kept and thus more DCS consensus reads are produced.Moreover,if an SSCS family is missing its partner in a generated DCS consensus(singleton SSCS family),it is also kept when it has at least two members.To evaluate the performance of these changes,we compared our approach for all four diluted samples with the original duplex approach,which has cut-off of three for SSCS family size and only employs DCS sequences for variant calling(DCS_only),and an approach utilizing only SSCS sequences for variant calling(SSCS_only).Figure 3 shows the performance results of these three approaches for HD753M1 and HD753M2 at different AF cut-offs(detailed values are summarized in Table S5).In general,the SSCS_only approach had the best sensitivity but a low specificity due to its high mean depth.Unsurprisingly,SSCS_only generated more false positives due to PCR-amplified errors or DNA damage in single-strand sequences.In contrast,DCS_only had fewer false positives,as DCS is much more accurate.However,a large amount of data were disregarded in the DCS_only approach,as some low-frequency true mutations were classified as false negatives due to low depth.By employing both DCS and SSCS sequences for variant calling,our SinoDuplex approach achieved a better balance of sensitivity and specificity.

    Figure 2 The correlation of detected AF with expected AF for mutations detected in four diluted samples

    Impact of sequencing depth and SSCS family size

    To determine the appropriate sequencing depth required for SinoDuplex to detect low-allele-frequency variant calling from cfDNA samples,we performed in silico downsampling for the sample HD753M2 at different depths from raw sequencing data and then calculated sensitivity and specificity after generating SSCS and DCS reads.Figure 4 shows that,as sequencing depth decreased,sensitivity dropped dramatically while specificity increased slightly at a limit of detection of 0.1%.The reason was that some rare mutations would not be detected at low depth and several other false positive variants appeared at higher sequencing depth.The best sensitivity and specificity results using SinoDuplex were achieved at a depth of 1831×.Obviously,there was an intriguing relationship between depth and family size.Therefore,we checked the distribution of SSCS family size with different sequencing depths.At a depth of 1968x,the family size of HD753M2 peaked expectably at 4,yielding high quality SSCS with optimal read numbers(Figure S3).However,a left shift of the peak generally occurred in the downsampling samples of lower depths.According to Equations(2)and(3),the SSCS family size itself also has an impact on the quality of consensus sequences.We counted the average base quality of the final consensus sequence and found a similar trend in peak family size.When increasing family size from 1 to 4, the corresponding base quality increased from 75 to 86.In contrast,lowering family size resulted in poorer base quality in the consensus sequences.In our experience,a reliable SSCS is generated when the read number in the family is about 3 to 6.More member reads in the same SSCS family would not contribute to the yield of SSCS but increase sequencing cost.

    Validation results with ddPCR

    Figure 3 Comparison of the performance of three different approaches at different AF cut-offs

    Figure 4 Impact of sequencing depth on the performance of SinoDuplex

    Although good performance was observed in the reference standard samples,we also assessed the reliability of SinoDuplex in clinical samples and confirmed low-frequency mutations with ddPCR.Using customized panels of different sizes adapted for the SinoDuplex approach,a number of hotspot actionable mutations were detected with low frequency in cfDNA samples from lung cancer patients enrolled in our collaborating hospitals.Detection of variant allele frequencies below 5%remains a challenge and a subset of these variants detected by SinoDuplex with frequencies between 0.1%and 5%were validated by ddPCR.As summarized in Table 2,five of six actionable mutations from patient cfDNA samples detected by SinoDuplex were confirmed to have similar AFs by ddPCR.One mutation,EGFR T790M,had a frequency of 0.1%in patient P3 and was reported as negative with a very weak signal,as 0.1%is the limit of detection of ddPCR.Overall,our approach was consistent with ddPCR for these lowfrequency hotspot actionable variants.

    Discussion

    ctDNA refers to the fraction of cfDNA in a patient’s blood that originates from a tumor.Noninvasive access to cancerderived DNA is particularly attractive for solid tumors,allowing repeated sampling without invasive procedures.Advances in DNA sequencing technologies and our understanding of tumor molecular biology have resulted in increased interest in exploiting ctDNA as a tool to facilitate earlier detection of cancer and thereby improve therapeutic outcomes byenabling early intervention[38].However,the application of this method has been challenging due to the low sensitivity in analyzing trace amounts of ctDNA in blood. Many sequencing technologies and algorithms have been developed and optimized to improve the detection accuracy of low frequency variants from ctDNA samples.One of these methods,DS technology,is particularly sensitive and uses attachment of unique molecular identifiers (UMI) to DNA templates to detect and quantify low-frequency genetic alterations in ctDNA samples [10,21]. Its power comes from pooling together multiple descendants of both strands of the original DNA molecules,allowing true variants to be distinguished from PCR amplification and sequencing artifacts.However,the method’s reliance on multiple sequencing reads of the same molecule means that DS requires much larger sequencing capacity than conventional NGS to produce a given depth of sequencing data,making it prohibitively expensive for broad usage in clinical settings.Furthermore,every duplex experiment produces a substantial proportion of singleton SSCS families that cannot be used in the analysis and are technically thrown away.Despite the great promise of DS,methods for both the experimental and computational aspects of this technique are still evolving.

    Table 2 Validation of hotspot actionable mutations detected by SinoDuplex with ddPCR assay

    To process cfDNA DS data more efficiently,we developed an improved duplex barcoding strategy and a novel computational analysis algorithm called SinoDuplex to generate lowerror consensus sequences from raw sequencing data.Taking advantage of degenerate UMIs,our duplex adapters provide significant advantages over the 12 N duplex adapters originally described[10,21]and the commercially available 3 N duplex adapter(Integrated DNA Technologies,Coralville,IA).First,SinoDuplex adapters are much more cost-effective than commercially available 3 N duplex adapters and can be easily acquired by annealing each pre-defined UMI pair and pooling them together.In contrast,synthesis of 12 N duplex adapters requires a series of enzymatic and purification steps.Second,SinoDuplex adapters are sufficient to identify most original nucleic acid molecules possessing the same genomic coordinates(start and end positions)in a sample while using far fewer barcode combinations compared to 12 N duplex adapters.Finally,pre-defined UMIs with at least three edit distances ensure improved accuracy for identification, while degenerate UMIs may be affected by single-base mismatches introduced by amplification or sequencing errors,leading to erroneous identification.Figure S4 shows a detailed graphical depiction of these different duplex adapters.Moreover,compared to DS analysis approaches reported previously[10,20,21], two major improvements were implemented in SinoDuplex to generate consensus sequences more precisely and utilize raw sequencing data more efficiently.The first one is the adoption of Bayesian theory[Equations(1),(2),and(3)]to generate SSCS sequences and quality scores.The other one is a lowering of the read cut-off from three to two in generating SSC sequences and retaining singleton SSCS families in subsequent analyses.The consensus sequence generation step significantly reduced PCR amplification errors and sequencing errors.Retaining singleton SSCS consensus reads improved utilization of raw sequencing data and reduced sequencing costs.SinoDuplex can be easily integrated into any existing pipelines that analyze DS data.

    Conclusions

    In this study,we present SinoDuplex,a promising DS method aimed at improving the sensitivity of detection of lowfrequency mutations in cfDNA.Compared to original DS approaches, our new computational analysis algorithm increased the yield of DS data while making it more costeffective.The method significantly improves the sensitivity and specificity in detecting extremely low frequency variants in plasma cfDNA samples.The potential of our approach for routine clinical applications will be of great importance for physicians,providing them with a powerful tool to diagnose tumors,monitor tumor dynamics,and evaluate patient responses to targeted therapy.

    Data availability

    The source code of SinoDuplex is freely available for academic use from https://github.com/SinOncology/sinoduplex. The raw duplex sequencing data reported in this paper have been deposited in the Genome Sequence Archive [39] at the National Genomics Data Center,Beijing Institute of Genomics,Chinese Academy of Sciences/China National Center for Bioinformation (GSA: CRA001933), and are publicly accessible at https://bigd.big.ac.cn/gsa.

    Authors’contributions

    BH conceived the idea and supervised the study.BH and YR designed the algorithm.YR and DW implemented the algorithm and performed the main analysis.YZ performed the laboratory validation experiments.FL,YF,SX,LS,JL,and HD analyzed the results and contributed to the manuscript writing.YR,YZ,DW drafted the manuscript,and BH edited the manuscript. ALL authors read and approved the final manuscript.

    Competing interests

    The authors have declared that no competing interests exist.

    Acknowledgments

    This work was financed by Grant-in-aid for scientific research from the Guangzhou Science and Technology Plan projects of China(Grant No.201802020004).

    Supplementary material

    Supplementary data to this article can be found online at https://doi.org/10.1016/j.gpb.2020.02.003.

    ORCID

    0000-0003-2132-252X(Ren Y)

    0000-0002-0098-8974(Zhang Y)

    0000-0003-2485-3019(Wang D)

    0000-0003-1896-963X(Liu F)

    0000-0001-5211-8930(Fu Y)

    0000-0001-6910-7322(Xiang S)

    0000-0002-5990-8457(Su L)

    0000-0003-2148-9872(Li J)

    0000-0001-8024-458X(Dai H)

    0000-0002-4748-2882(Huang B)

    国产精品1区2区在线观看.| 国产精品精品国产色婷婷| 超碰av人人做人人爽久久| 欧美色视频一区免费| 男人狂女人下面高潮的视频| 国产伦精品一区二区三区四那| 一个人看视频在线观看www免费| 午夜福利在线观看免费完整高清在 | 午夜福利视频1000在线观看| 日韩欧美在线二视频| 中文字幕av在线有码专区| 亚洲专区国产一区二区| 国产 一区精品| 3wmmmm亚洲av在线观看| 成年女人永久免费观看视频| 丰满人妻一区二区三区视频av| 亚洲精品成人久久久久久| 男人舔女人下体高潮全视频| 欧美性猛交╳xxx乱大交人| 最近在线观看免费完整版| 一卡2卡三卡四卡精品乱码亚洲| 中国美白少妇内射xxxbb| 看十八女毛片水多多多| ponron亚洲| 色吧在线观看| 日韩av在线大香蕉| 老司机午夜福利在线观看视频| 99久久成人亚洲精品观看| 精品久久久久久久久久久久久| 99久国产av精品| 欧美成人免费av一区二区三区| 校园春色视频在线观看| 又黄又爽又刺激的免费视频.| 在线免费十八禁| 国产亚洲精品久久久com| 成人无遮挡网站| 97热精品久久久久久| 毛片一级片免费看久久久久 | 麻豆成人午夜福利视频| 国内精品宾馆在线| 国产淫片久久久久久久久| 可以在线观看的亚洲视频| 国产精品美女特级片免费视频播放器| av女优亚洲男人天堂| 国产亚洲av嫩草精品影院| 免费搜索国产男女视频| 麻豆成人午夜福利视频| 联通29元200g的流量卡| .国产精品久久| 亚洲成人精品中文字幕电影| 舔av片在线| 国产成人av教育| 啦啦啦观看免费观看视频高清| 亚洲人成网站在线播放欧美日韩| 黄色女人牲交| 男女视频在线观看网站免费| 欧美日韩亚洲国产一区二区在线观看| 日韩,欧美,国产一区二区三区 | 亚洲av电影不卡..在线观看| 亚洲国产精品合色在线| 男人舔奶头视频| 欧美一区二区国产精品久久精品| 国产麻豆成人av免费视频| av专区在线播放| 我的老师免费观看完整版| 精品午夜福利在线看| 一本久久中文字幕| 日韩在线高清观看一区二区三区 | 成人高潮视频无遮挡免费网站| 午夜福利在线在线| 少妇裸体淫交视频免费看高清| 免费在线观看成人毛片| 国产精品不卡视频一区二区| 亚洲美女黄片视频| 久久久久久久精品吃奶| 日韩欧美在线二视频| 白带黄色成豆腐渣| 成人国产麻豆网| 亚洲av熟女| 天天躁日日操中文字幕| 可以在线观看的亚洲视频| 99热这里只有是精品50| 午夜福利高清视频| 18禁在线播放成人免费| 蜜桃亚洲精品一区二区三区| 看十八女毛片水多多多| 99久久中文字幕三级久久日本| 久久久久国内视频| 亚洲在线观看片| 1000部很黄的大片| a级毛片a级免费在线| 亚洲欧美激情综合另类| 日韩国内少妇激情av| 精品久久久久久久久亚洲 | 看免费成人av毛片| 成人国产一区最新在线观看| 18禁在线播放成人免费| 亚洲黑人精品在线| 麻豆成人午夜福利视频| 免费在线观看日本一区| 国产高清不卡午夜福利| 亚洲自偷自拍三级| 午夜福利在线观看吧| 一级av片app| 色综合婷婷激情| 1000部很黄的大片| 极品教师在线免费播放| 国产精品美女特级片免费视频播放器| 免费人成在线观看视频色| .国产精品久久| 国产精品福利在线免费观看| 露出奶头的视频| 日本欧美国产在线视频| 可以在线观看毛片的网站| 小说图片视频综合网站| 久久精品国产亚洲av香蕉五月| 国产精品日韩av在线免费观看| 18禁裸乳无遮挡免费网站照片| 欧美成人免费av一区二区三区| 日韩欧美在线乱码| 日本黄大片高清| 一区福利在线观看| 日本一二三区视频观看| 黄色丝袜av网址大全| 午夜视频国产福利| 国内精品美女久久久久久| 九九久久精品国产亚洲av麻豆| 91在线精品国自产拍蜜月| 三级毛片av免费| 国产高清三级在线| 国产乱人伦免费视频| 国产精品一及| 91久久精品国产一区二区成人| 最新在线观看一区二区三区| 国内久久婷婷六月综合欲色啪| 国产精品国产高清国产av| 久久精品国产清高在天天线| 久久天躁狠狠躁夜夜2o2o| 午夜福利成人在线免费观看| 欧美绝顶高潮抽搐喷水| 亚洲乱码一区二区免费版| 麻豆国产97在线/欧美| 中亚洲国语对白在线视频| 欧美中文日本在线观看视频| 免费无遮挡裸体视频| 久久精品久久久久久噜噜老黄 | 亚洲av免费高清在线观看| 九九爱精品视频在线观看| 老司机深夜福利视频在线观看| 日日啪夜夜撸| 亚洲精品国产成人久久av| 好男人在线观看高清免费视频| 国产大屁股一区二区在线视频| 少妇高潮的动态图| 亚洲乱码一区二区免费版| 91午夜精品亚洲一区二区三区 | 免费av毛片视频| 亚洲aⅴ乱码一区二区在线播放| 亚洲人成伊人成综合网2020| 亚洲三级黄色毛片| 色哟哟哟哟哟哟| 国产精品久久久久久精品电影| 亚洲国产色片| 欧美一级a爱片免费观看看| 麻豆成人av在线观看| 久久久成人免费电影| 99国产精品一区二区蜜桃av| 成年女人毛片免费观看观看9| 欧美+日韩+精品| 久久香蕉精品热| 欧美精品国产亚洲| 99精品久久久久人妻精品| 欧美一级a爱片免费观看看| 国产一级毛片七仙女欲春2| 精品一区二区免费观看| 色精品久久人妻99蜜桃| 午夜久久久久精精品| 国内精品一区二区在线观看| 亚洲精品色激情综合| 午夜福利视频1000在线观看| 老女人水多毛片| 国产视频一区二区在线看| 日韩欧美一区二区三区在线观看| 18禁裸乳无遮挡免费网站照片| av黄色大香蕉| 在线天堂最新版资源| 热99在线观看视频| 久久久国产成人免费| 成人综合一区亚洲| 12—13女人毛片做爰片一| 欧美精品国产亚洲| 欧美xxxx性猛交bbbb| 成人欧美大片| 国产精品一区二区性色av| 午夜福利18| 欧美一区二区国产精品久久精品| 韩国av在线不卡| 国产亚洲精品综合一区在线观看| 国产精品久久久久久精品电影| 午夜福利在线观看吧| 亚洲人成网站高清观看| 国产女主播在线喷水免费视频网站 | 亚洲最大成人av| bbb黄色大片| 成人一区二区视频在线观看| 午夜福利在线观看免费完整高清在 | 国产色婷婷99| 一区二区三区高清视频在线| or卡值多少钱| 成人鲁丝片一二三区免费| 亚洲va日本ⅴa欧美va伊人久久| 五月伊人婷婷丁香| 天天躁日日操中文字幕| 欧美日本视频| 国产 一区精品| 日本a在线网址| 一级av片app| 人妻制服诱惑在线中文字幕| 久久久久久伊人网av| 精品久久久久久久末码| 国产精品不卡视频一区二区| 国产一区二区三区av在线 | 又紧又爽又黄一区二区| 综合色av麻豆| 欧美3d第一页| 日韩精品中文字幕看吧| av在线观看视频网站免费| 亚洲自偷自拍三级| 乱系列少妇在线播放| 欧美不卡视频在线免费观看| 精品国产三级普通话版| 97热精品久久久久久| 亚洲美女搞黄在线观看 | 久久久久免费精品人妻一区二区| 精品人妻1区二区| 国产成人影院久久av| 久久人妻av系列| 国内少妇人妻偷人精品xxx网站| 丰满乱子伦码专区| 蜜桃亚洲精品一区二区三区| 欧美最新免费一区二区三区| 波多野结衣高清作品| 亚洲成人久久爱视频| 亚洲国产日韩欧美精品在线观看| 少妇人妻一区二区三区视频| 搡老妇女老女人老熟妇| 久久精品国产亚洲av天美| 一区二区三区激情视频| 哪里可以看免费的av片| 久久精品国产亚洲网站| 免费看光身美女| 亚洲精品成人久久久久久| 少妇猛男粗大的猛烈进出视频 | 在线观看舔阴道视频| 校园人妻丝袜中文字幕| 中国美女看黄片| 久久精品国产亚洲网站| 美女cb高潮喷水在线观看| 久久午夜福利片| 给我免费播放毛片高清在线观看| 老司机福利观看| 免费在线观看日本一区| 日韩精品中文字幕看吧| 欧美又色又爽又黄视频| 91麻豆精品激情在线观看国产| 两个人视频免费观看高清| 久久天躁狠狠躁夜夜2o2o| 国产日本99.免费观看| 国产精品一区二区免费欧美| 欧美人与善性xxx| 亚洲电影在线观看av| 亚洲精品在线观看二区| 亚洲av.av天堂| 我要搜黄色片| 亚洲四区av| 亚洲av不卡在线观看| 国产探花在线观看一区二区| 国产精品人妻久久久久久| 小说图片视频综合网站| 亚洲精品成人久久久久久| 久久久久久久午夜电影| 在线免费观看不下载黄p国产 | 国产蜜桃级精品一区二区三区| 亚洲最大成人手机在线| 日韩在线高清观看一区二区三区 | 国产私拍福利视频在线观看| 精品一区二区三区视频在线| 欧美日韩国产亚洲二区| 丰满的人妻完整版| 可以在线观看的亚洲视频| 久久香蕉精品热| 国产欧美日韩精品亚洲av| 午夜福利在线观看吧| 免费观看的影片在线观看| 精品人妻1区二区| 九九热线精品视视频播放| 亚洲av一区综合| 美女cb高潮喷水在线观看| 欧美中文日本在线观看视频| 成人欧美大片| 国产一区二区三区视频了| 国产精品国产三级国产av玫瑰| 欧美成人一区二区免费高清观看| 亚洲天堂国产精品一区在线| 日韩中文字幕欧美一区二区| 国产午夜精品久久久久久一区二区三区 | 日韩欧美国产一区二区入口| 久久精品夜夜夜夜夜久久蜜豆| 久久精品久久久久久噜噜老黄 | 女同久久另类99精品国产91| 欧美日本亚洲视频在线播放| 亚洲熟妇中文字幕五十中出| 午夜老司机福利剧场| 男人狂女人下面高潮的视频| 少妇的逼水好多| 国内精品美女久久久久久| 国产精品一及| 欧美xxxx黑人xx丫x性爽| 免费一级毛片在线播放高清视频| 午夜a级毛片| 午夜精品久久久久久毛片777| 91狼人影院| 亚洲综合色惰| 直男gayav资源| 久久久久久九九精品二区国产| 观看美女的网站| 日本一本二区三区精品| 一级a爱片免费观看的视频| 高清毛片免费观看视频网站| 在线观看舔阴道视频| 男女做爰动态图高潮gif福利片| 亚洲四区av| 国产精品久久久久久久电影| 一级毛片久久久久久久久女| 国产在线精品亚洲第一网站| 无遮挡黄片免费观看| 三级毛片av免费| 直男gayav资源| 女同久久另类99精品国产91| 五月玫瑰六月丁香| 欧美3d第一页| 国产免费男女视频| 91av网一区二区| 亚洲黑人精品在线| 中文字幕久久专区| 久久精品国产自在天天线| 亚洲欧美日韩高清专用| 欧美+日韩+精品| 欧美一区二区亚洲| 欧美一级a爱片免费观看看| 精品久久久久久,| 午夜老司机福利剧场| 国产精品人妻久久久久久| 香蕉av资源在线| 桃色一区二区三区在线观看| 永久网站在线| 久久久国产成人免费| 国产精品98久久久久久宅男小说| 国产亚洲91精品色在线| 九九爱精品视频在线观看| 亚洲专区中文字幕在线| 夜夜爽天天搞| 波多野结衣巨乳人妻| 婷婷六月久久综合丁香| 久久99热6这里只有精品| 国产三级在线视频| 能在线免费观看的黄片| av福利片在线观看| 偷拍熟女少妇极品色| 亚洲三级黄色毛片| 午夜a级毛片| 内地一区二区视频在线| 午夜福利18| 亚洲av免费高清在线观看| 看免费成人av毛片| 深爱激情五月婷婷| 天天一区二区日本电影三级| 变态另类丝袜制服| 精品久久久久久久久久久久久| 国产女主播在线喷水免费视频网站 | 成年免费大片在线观看| 身体一侧抽搐| 熟女人妻精品中文字幕| 在线a可以看的网站| 搡老妇女老女人老熟妇| 99久久精品一区二区三区| 深夜精品福利| 高清在线国产一区| 97人妻精品一区二区三区麻豆| 99精品久久久久人妻精品| 三级男女做爰猛烈吃奶摸视频| 美女xxoo啪啪120秒动态图| av在线老鸭窝| 久久中文看片网| 一卡2卡三卡四卡精品乱码亚洲| 欧美又色又爽又黄视频| 成人性生交大片免费视频hd| 亚洲电影在线观看av| 日韩亚洲欧美综合| 在线a可以看的网站| 日本免费a在线| 国产av麻豆久久久久久久| 国产精品免费一区二区三区在线| 亚洲avbb在线观看| 精品人妻1区二区| 最新在线观看一区二区三区| 亚洲精品亚洲一区二区| 欧美色视频一区免费| 欧美一区二区精品小视频在线| 亚洲精品影视一区二区三区av| 国产精品1区2区在线观看.| 18禁裸乳无遮挡免费网站照片| 国产主播在线观看一区二区| 干丝袜人妻中文字幕| 国产高潮美女av| 成人av在线播放网站| 欧美三级亚洲精品| 国产精品美女特级片免费视频播放器| 国产一级毛片七仙女欲春2| 性插视频无遮挡在线免费观看| 99久久精品一区二区三区| 国产成年人精品一区二区| 蜜桃久久精品国产亚洲av| 色综合亚洲欧美另类图片| 亚洲男人的天堂狠狠| 深夜a级毛片| 久99久视频精品免费| 嫩草影视91久久| 久久久国产成人免费| 国产毛片a区久久久久| 夜夜爽天天搞| 欧美最新免费一区二区三区| 国产亚洲精品久久久com| 女同久久另类99精品国产91| 少妇猛男粗大的猛烈进出视频 | 日日啪夜夜撸| 最新在线观看一区二区三区| 亚洲精品成人久久久久久| 日韩欧美在线二视频| 精品一区二区免费观看| 最新在线观看一区二区三区| 久久精品久久久久久噜噜老黄 | 婷婷亚洲欧美| 免费看美女性在线毛片视频| 亚洲va日本ⅴa欧美va伊人久久| 淫秽高清视频在线观看| 亚洲av成人精品一区久久| 麻豆国产97在线/欧美| 99精品在免费线老司机午夜| 黄色视频,在线免费观看| 国产欧美日韩精品一区二区| 日韩欧美一区二区三区在线观看| 亚洲在线观看片| 婷婷六月久久综合丁香| 非洲黑人性xxxx精品又粗又长| 日本-黄色视频高清免费观看| 国产精品美女特级片免费视频播放器| 深夜a级毛片| 别揉我奶头 嗯啊视频| 国产精品一区二区三区四区久久| 九色国产91popny在线| 亚洲精品在线观看二区| 黄色视频,在线免费观看| 国产av不卡久久| 欧美色视频一区免费| 欧美人与善性xxx| avwww免费| 夜夜夜夜夜久久久久| .国产精品久久| a级毛片免费高清观看在线播放| 亚洲成人精品中文字幕电影| 中亚洲国语对白在线视频| 18禁黄网站禁片午夜丰满| 在线观看舔阴道视频| 黄色女人牲交| 51国产日韩欧美| 真实男女啪啪啪动态图| 国产三级中文精品| 制服丝袜大香蕉在线| 99久久成人亚洲精品观看| 亚洲精品亚洲一区二区| 一级毛片久久久久久久久女| 午夜福利成人在线免费观看| 两个人的视频大全免费| 国产黄色小视频在线观看| 如何舔出高潮| 波野结衣二区三区在线| 国国产精品蜜臀av免费| 国产av一区在线观看免费| 日本黄大片高清| 国模一区二区三区四区视频| 久久久久国产精品人妻aⅴ院| 日本 欧美在线| 国产精品1区2区在线观看.| 色播亚洲综合网| 亚洲经典国产精华液单| 毛片一级片免费看久久久久 | 欧美人与善性xxx| 成人特级av手机在线观看| 日韩一区二区视频免费看| 国产一区二区在线av高清观看| 精品国内亚洲2022精品成人| 欧美精品国产亚洲| 窝窝影院91人妻| 99久久中文字幕三级久久日本| 亚洲不卡免费看| 男女做爰动态图高潮gif福利片| 欧洲精品卡2卡3卡4卡5卡区| 久久久久久伊人网av| 啦啦啦韩国在线观看视频| 久久精品综合一区二区三区| 日韩一区二区视频免费看| 男女边吃奶边做爰视频| 国产激情偷乱视频一区二区| 久久久久久大精品| 精品国内亚洲2022精品成人| 亚洲av一区综合| 久久精品国产亚洲av涩爱 | 欧美最黄视频在线播放免费| 丰满人妻一区二区三区视频av| 在线播放无遮挡| 国产精品女同一区二区软件 | 黄色视频,在线免费观看| 免费av观看视频| 超碰av人人做人人爽久久| 久久香蕉精品热| 久久九九热精品免费| 欧美三级亚洲精品| 99在线视频只有这里精品首页| 日韩人妻高清精品专区| 国产精品美女特级片免费视频播放器| 一a级毛片在线观看| 别揉我奶头 嗯啊视频| 香蕉av资源在线| 亚洲人成网站高清观看| 亚洲一级一片aⅴ在线观看| 国产午夜精品论理片| 日日干狠狠操夜夜爽| 亚洲va日本ⅴa欧美va伊人久久| 亚洲欧美日韩卡通动漫| 国产女主播在线喷水免费视频网站 | 日本五十路高清| 亚洲四区av| 91av网一区二区| 国产精品久久电影中文字幕| 国产私拍福利视频在线观看| 精品一区二区三区av网在线观看| 国产成人福利小说| 日日摸夜夜添夜夜添av毛片 | 真人一进一出gif抽搐免费| av中文乱码字幕在线| 国产69精品久久久久777片| 精品久久久久久久久亚洲 | 国产午夜精品久久久久久一区二区三区 | 久久人人爽人人爽人人片va| 免费观看精品视频网站| a级毛片免费高清观看在线播放| 久久久久免费精品人妻一区二区| 搡老妇女老女人老熟妇| 黄色配什么色好看| 最近中文字幕高清免费大全6 | 久久精品久久久久久噜噜老黄 | 一区二区三区四区激情视频 | 日韩中文字幕欧美一区二区| 国产单亲对白刺激| 国产亚洲91精品色在线| 亚洲av五月六月丁香网| 久久午夜福利片| 2021天堂中文幕一二区在线观| 国产精品一区二区免费欧美| 桃红色精品国产亚洲av| 色哟哟哟哟哟哟| 欧美xxxx性猛交bbbb| 亚洲精品一区av在线观看| 精品久久久久久久久亚洲 | 免费在线观看日本一区| 午夜福利欧美成人| 免费看日本二区| 亚洲自拍偷在线| 51国产日韩欧美| 亚洲电影在线观看av| 国产欧美日韩一区二区精品| 99久久久亚洲精品蜜臀av| 亚洲人与动物交配视频| 97碰自拍视频| 午夜福利在线观看吧| 12—13女人毛片做爰片一| 一级黄片播放器| 国产精品久久久久久亚洲av鲁大| 成人亚洲精品av一区二区| 国产精品国产三级国产av玫瑰| 国产在线精品亚洲第一网站| 亚洲熟妇熟女久久| 在线观看66精品国产| 国产精华一区二区三区| 国产爱豆传媒在线观看| 一进一出抽搐动态| 97人妻精品一区二区三区麻豆| 尤物成人国产欧美一区二区三区| 十八禁网站免费在线| 欧美日本视频| 性插视频无遮挡在线免费观看| 1000部很黄的大片| 久久99热6这里只有精品| 亚洲美女视频黄频| 国产探花在线观看一区二区| 九色成人免费人妻av| 两性午夜刺激爽爽歪歪视频在线观看| 久久午夜福利片| 两个人的视频大全免费| 啦啦啦啦在线视频资源| 麻豆国产97在线/欧美| 日韩国内少妇激情av| 午夜免费激情av| 中文字幕av成人在线电影|