Zubair Khalid

Virologist/Molecular Biologist | Veterinarian | Bioinformatician

Conventional & Molecular Virology • Vaccine Development • Computational Biology

Dr. Zubair Khalid is a veterinarian and virologist specializing in conventional and molecular virology, vaccine development, and computational biology. Dedicated to advancing animal health through innovative research and multi-omics approaches.

Dr. Zubair Khalid - Veterinarian, Virologist, and Vaccine Development Researcher specializing in Computational Biology, Multi-omics, Animal Health, and Infectious Disease Research

Category: Guides

Bgi Genomics

BGI Genomics is a global biotechnology company that provides high throughput genome sequencing, genotyping, and multi omics services. This guide is intended for researchers, clinicians, and laboratory professionals who want an evidence based overview of BGI’s core technologies, a structured approach to selecting and executing a sequencing project, and a realistic understanding of the technical limits and common pitfalls encountered when working with BGI data. The information here is drawn from publicly available technical references and recent peer reviewed studies that use BGI sequencing platforms or related methods NCBI Bookshelf. It is not a substitute for your own validation and local quality control.

At a Glance

Aspect Key Details
Core technology DNA nanoball (DNB) sequencing, combinatorial probe anchor synthesis (cPAS)
Primary platforms DNBSEQ T7, DNBSEQ G400, DNBSEQ MGI (formerly BGISEQ)
Typical read lengths Short reads (100-150 bp paired end, also 50 bp single end for some applications)
Typical data output per run Up to 6 Tb (T7), 60-1440 Gb (G400) depending on flow cell
Major applications Whole genome sequencing, whole exome sequencing, RNA seq, single cell sequencing, spatial transcriptomics, non invasive prenatal testing (NIPT)
Data formats produced FASTQ, BCL (raw), with downstream processing to BAM, VCF etc.
Public data repositories NCBI Sequence Read Archive (SRA), EBI ENA, CNGBdb
Quality metrics Q30 ≥ 85% (typical), duplication rate, mapping rate, coverage uniformity

Core Concepts and Technology

BGI Genomics builds its sequencing on the DNB (DNA nanoball) technology, which is a fundamentally different approach from the clonal bridge amplification used by Illumina. DNB technology circularizes library fragments, amplifies them in a single solution using rolling circle amplification, and then loads the resulting nanoballs onto a patterned array flow cell. Sequencing by synthesis uses cPAS (combinatorial probe anchor synthesis) to read bases one by one. The key advantage of DNB is that the nanoballs do not grow on the flow cell surface, they are deposited at high density and do not form clusters that can collapse or dephase. This design reduces error accumulation and often yields lower duplication rates compared to conventional bridge amplification NCBI Sequence Read Archive.

BGI’s platforms are widely used in large cohort studies and clinical applications. For example, the SpatioTemporal Omics Consortium uses BGI sequencing and stereomics to map gene expression across tissues and time, providing a global reference for biological discovery SpatioTemporal Omics Consortium. The technology also supports non invasive prenatal paternity testing for trisomy 21 fetuses, leveraging allele counting from maternal blood samples Allele counting based non invasive prenatal paternity testing. Understanding these core principles helps you evaluate whether BGI sequencing fits your project’s throughput, read length, and accuracy requirements.

Decision Criteria for Choosing BGI Services

Deciding to use BGI Genomics involves evaluating several practical factors:

  • Data volume per project. BGI’s high output platforms (e.g., T7) can deliver multiple terabases in a single run, making them cost effective for population scale whole genome sequencing. For smaller targeted panels, a lower output flow cell (G400) may be more economical.
  • Read length constraints. BGI typically produces short reads (up to 150 bp paired end). If your application needs long reads for de novo assembly or structural variant detection, BGI offers a long read option (single molecule sequencing) but most standard services are short read. Check with your provider for current offerings.
  • Library compatibility. BGI uses specific adapters and requires libraries constructed according to their protocols. If you already have Illumina compatible libraries, you cannot simply sequence them on a BGI machine without a conversion step (adapter swapping or re ligation). Plan for additional cost and time.
  • Bioinformatics ecosystem. BGI’s proprietary base calling software produces FASTQ files that are largely compatible with standard bioinformatics pipelines. However, some downstream tools (especially those designed for Illumina specific error models) may need parameter adjustments. The Galaxy Training Network offers tutorials for processing both BGI and Illumina data Galaxy Training Network.
  • Regulatory and data privacy. For clinical applications, verify that the BGI service lab you plan to use holds the appropriate certifications (e.g., CLIA, CAP, ISO) and complies with local data protection regulations. Many BGI facilities are based in China, which has different privacy laws than the EU or US.

Practical Workflow for BGI Sequencing Projects

A well defined project workflow reduces mistakes and ensures reproducible results. The steps below outline a typical BGI sequencing project from sample to final analysis.

  1. Project design and library preparation. Define your biological question and required coverage. Prepare DNA or RNA libraries using BGI recommended kits or validated third party kits. For RNA seq, include polyA selection or ribodepletion. For single cell experiments, use appropriate 10x Genomics or BGI compatible single cell prep.
  2. Quality control of libraries. Assess concentration, fragment size distribution, and adapter contamination. Use a Bioanalyzer or TapeStation. BGI’s library QC thresholds are similar to those of other platforms: a tight size distribution and minimal free adapters.
  3. Pooling and loading. If you are multiplexing, pool libraries in equimolar amounts. BGI’s flow cell loading uses a DNB loader that deposits nanoballs onto the patterned array. Overloading can cause overlapping signals and reduced data quality.
  4. Sequencing run. The run is managed by BGI’s sequencing software. Monitor real time metrics such as signal intensity, error rate per cycle, and cluster density. Typical run duration for 150 PE on a T7 is about 40 hours.
  5. Base calling and output. BGI’s base caller (e.g., ZEBRA) converts raw images to FASTQ files. Download the output files from the instrument or via BGI’s cloud platform.
  6. Primary QC of FASTQ files. Use FastQC, MultiQC, or similar tools to check per base quality, GC content, adapter content, and duplication. Compare against your project’s threshold (e.g., Q30 > 85% for most bases). EMBL EBI Training provides guidelines for assessing sequencing data quality EMBL EBI Training.
  7. Alignment and variant calling. Map reads to a reference genome (e.g., GRCh38 for human) using a short read aligner such as BWA MEM or Bowtie2. BGI’s own pipeline uses a custom aligner, but standard open source tools work well. Call variants with GATK, FreeBayes, or Strelka. For RNA seq, use STAR or HISAT2. For single cell data, process with Cell Ranger (if compatible) or a BGI equivalent.
  8. Downstream analysis. Perform differential expression, pathway analysis, or copy number calling as needed. For spatial transcriptomics, tools like Seurat or Squidpy can integrate BGI stereo seq data. A recent study using BGI spatial transcriptomics uncovered vasculature centered cellular interactions driving Japanese encephalitis progression in a mouse model, illustrating the power of spatially resolved expression data Spatial transcriptomics uncovers vasculature centered interactions.
  9. Data archiving. Submit raw data to a public repository if required by your funding agency or journal. The NCBI Sequence Read Archive (SRA) accepts BGI generated FASTQ files under the same BioProject structure as other platforms.

Quality Checks and Validation

Quality control should be applied at every stage of the workflow. Here are specific checks for BGI data:

  • Base quality distribution. BGI data typically shows a slight dip in Q scores in the first few cycles (similar to Illumina) and a gradual decline toward the end of the read. A sharp drop may indicate loading issues.
  • Duplication rate. Because DNB nanoballs are monoclonal and deposited separately, BGI data often exhibits lower duplication rates than Illumina. A high duplication rate (e.g., > 30% in a whole genome dataset) may indicate over amplification during library prep rather than a sequencing artifact.
  • GC bias. BGI sequencing can show moderate GC bias, especially in regions with extreme GC content. Compare the observed GC distribution to the expected distribution from the reference genome. Tools like Picard CollectGcBiasMetrics can quantify this.
  • Mapping rate. Expect > 90% of reads to map to the reference genome in a typical human WGS project. Lower mapping rates may indicate contamination or adapter issues.
  • Cross contamination. Use a tool such as ContEst to check for sample swaps. For clinical samples, also check for fetal fraction in NIPT or spike in controls.

Common Mistakes and Pitfalls

  • Assuming read length is always sufficient. For metagenomics or splice variant analysis, 150 bp paired end may not resolve complex repetitive regions. Plan your read length early.
  • Neglecting adapter compatibility. Many commercial library prep kits are designed for Illumina adapters. BGI uses different adapter sequences, if you use a third party kit, verify compatibility or be prepared to swap adapters before loading.
  • Underestimating computational requirements. BGI’s high output runs produce enormous amounts of data. A single T7 run can generate 6 Tb of raw data, which requires significant storage, RAM, and CPU time for alignment and variant calling. Use cloud resources or local high performance clusters.
  • Relying solely on BGI’s primary analysis. The supplied FASTQ files are usually of good quality, but always run your own independent QC. Platform specific base callers may not flag all artifacts.
  • Ignoring batch effects. If you are running samples across multiple flow cells or different BGI platforms (e.g., some on G400 and some on T7), include technical replicates and use normalization methods to correct for batch effects.
  • Overlooking the need for context aware parameter tuning. For long read mapping or specialized analyses like structural variant detection, simulation based optimization of mapping parameters can dramatically improve accuracy. A recent study demonstrated that context aware simulation enables systematic optimization of long read mapping parameters, a principle that also applies to short read data from BGI Context aware simulation enables systematic optimization.

Limits of Interpretation and Uncertainty

Any genomic result derived from BGI data carries inherent uncertainty that must be communicated clearly.

  • Base calling errors at homopolymer regions. DNB sequencing can produce insertion/deletion errors in long homopolymer runs, similar to other amplification based methods. Variants called in such regions should be filtered by homopolymer length or validated by an orthogonal method.
  • Coverage depth variations. Even with uniform loading, coverage is not perfectly uniform across the genome. Regions with extreme GC content, segmental duplications, or centromeres will have reduced coverage. This affects copy number calling and variant detection in those regions.
  • Allele dropout. In NIPT and other low frequency variant detection, a small proportion of fetal or somatic variants may not be detected due to stochastic sampling. Bayesian or empirical false discovery rate methods can help set appropriate detection thresholds.
  • Reference genome bias. Aligning to a single reference genome may miss population specific variants or cause mapping artifacts in divergent regions. Consider using a pangenome graph or decoy sequences for diverse populations.
  • Interpretation of functional variants. A variant identified from BGI sequencing may be a true somatic mutation or a germline polymorphism. Without matched normal samples or population databases (gnomAD, etc.), functional interpretation is speculative. The literature on gene expression profiling and cortical layer parcellation shows that even well validated expression data requires careful statistical adjustment for cell type composition Gene expression profiling enables refined parcellation.

Frequently Asked Questions

1. Can I sequence Illumina prepared libraries on a BGI machine directly? No, you cannot. Illumina libraries use different adapter sequences and cluster amplification chemistry. You must perform an adapter conversion step or prepare new libraries using BGI compatible kits. Some service providers offer library conversion as a paid add on.

2. How does the quality of BGI data compare to Illumina data? For most short read applications, the quality is comparable. BGI data typically shows lower duplication rates and similar Q30 scores. However, error modes differ slightly (e.g., more homopolymer indels in BGI). You should run the same bioinformatics pipeline but validate a few key variant calls with an orthogonal method, especially for clinical use.

3. Is BGI sequencing suitable for single cell transcriptomics? Yes, BGI offers dedicated single cell sequencing services and compatible library preparation kits. Many single cell analysis workflows (e.g., Seurat, Scanpy) accept BGI generated FASTQ files. Check the documentation for your preferred tool to confirm that the UMI and barcode structures are correctly recognized.

4. Where can I access public BGI sequencing data for benchmarking? The NCBI Sequence Read Archive (SRA) and the China National GeneBank DataBase (CNGBdb) host thousands of BGI generated datasets. You can search by the sequencing platform (e.g., “DNBSEQ” or “BGISEQ”) to find examples for testing pipelines.

References and Further Reading

Related Articles