SISSOR (Single-Stranded Sequencing using micrOfluidic Reactors)

About

See the paper here: http://www.pnas.org/content/114/47/12512.short

Abstract (from PNAS): Ultra-Accurate Genome Sequencing And Haplotyping Of Single Human Cells

Accurate detection of variants and long-range haplotypes in genomes of single human cells remains very challenging. Common approaches require extensive in vitro amplification of genomes of individual cells using DNA polymerases and high-throughput short-read DNA sequencing. These approaches have two notable drawbacks. First, polymerase replication errors could generate tens of thousands of false-positive calls per genome. Second, relatively short sequence reads contain little to no haplotype information. Here we report a method, which is dubbed SISSOR (single-stranded sequencing using microfluidic reactors), for accurate single-cell genome sequencing and haplotyping. A microfluidic processor is used to separate the Watson and Crick strands of the double-stranded chromosomal DNA in a single cell and to randomly partition megabase-size DNA strands into multiple nanoliter compartments for amplification and construction of barcoded libraries for sequencing. The separation and partitioning of large single-stranded DNA fragments of the homologous chromosome pairs allows for the independent sequencing of each of the complementary and homologous strands. This enables the assembly of long haplotypes and reduction of sequence errors by using the redundant sequence information and haplotype-based error removal. We demonstrated the ability to sequence single-cell genomes with error rates as low as 10^−8 and average 500-kb-long DNA fragments that can be assembled into haplotype contigs with N50 greater than 7 Mb. The performance could be further improved with more uniform amplification and more accurate sequence alignment. The ability to obtain accurate genome sequences and haplotype information from single cells will enable applications of genome sequencing for diverse clinical needs.

UPDATE 4/10/19: SISSOR phased VCF Data Release

A phased VCF containing the haplotypes assembled using the 3 single cells is available here: SISSOR Phased VCF

UPDATE 1/15/18: SISSOR Strand Information Data Release

For easy inspection of the data without running the pipeline, we have converted the SISSOR strand information to a set of GFF files. These files contain all of the reference and variant calls that were used for the haplotype-based strand-to-strand call validation. Also included is a bed-format file that describes the boundaries and haplotype assignments for each distinct haploid fragment (after haplotype processing). See the README.txt included in the tar archive for more information.

The data is available here: SISSOR Strand Data

SISSOR Computational Methods

This repository contains the majority of the codebase for analysis of SISSOR libraries and reproducing the analyses described in "Ultra-Accurate Genome Sequencing And Haplotyping Of Single Human Cells" (with the exception of the read mapping and the HMM-based fragment boundary segmentation steps).

It is important to note that the code is here for documentation purposes and has relatively little applicability except with data from the SISSOR device. Various components may be adapted in the future for general haplotype assembly or MDA-based protocols.

The general workflow, executed by the rules in the Snakefile, is the following:

1. Generate a high-quality BAC reference haplotype from BAC haplotype fragments for PGP1.
2. Generate a high-quality reference dataset of SNV and reference calls for PGP1 by intersecting high-quality SNV/ref datasets from two cell lines (lymphocyte complete genomics, and fibroblast 60x Illumina WGS).
3. Perform a samtools mpileup of 72 SISSOR libraries from PGP1f cells (3 cells x 24 reaction chambers).
4. Estimate various model parameters from the mpileup data.
5. Use custom variant caller (codename SISSORhands) to make SNV/ref calls at every genomic position in every SISSOR library, as well as consensus calls over all libraries.
6. Generate HapCUT2 format haplotype fragments from the SISSORhands variant calls, based on heterozygous SNV positions from 60x WGS of PGP1f cells.
7. Process haplotype fragments to split and/or filter them at (a) clusters of mixed alleles (haploid strands from two haplotypes overlapping in the same chamber to make a diploid mixture) and (b) detectable fragment-fragment switch errors (2 or more switches with respect to other fragments)
8. Assemble the processed haplotype fragments into haplotype blocks with HapCUT2.
9. Calculate accuracy of haplotype blocks with respect to BAC haplotypes from step [1] and also statistics such as N50.
10. Assign haplotype fragments back to haplotype blocks.
11. Calculate the accuracy of various base calling approaches with respect to reference set from step [2]:
    1. consensus calls from step [5]
    2. chamber-specific calls from step [5] that match between strands on the same haplotype:
       1. in the same cell (to get validated cell specific mutations)
       2. in all cells (to get maximum coverage of haplotype-validated calls)
       3. between different cells (cross-cell) (to estimate maximum bound on error rate of the haplotype-strand-matching method against a reference without inflation from cell-specific mutations)

For more details, see the SISSOR manuscript (above) and supplementary materials.

Requirements

• Snakemake (execution on a cluster, with appropriate cluster configuration, is basically required)
• BAM files for 3 SISSOR libraries, placed in bams/{cells}/ch{ch}.bam
  • {cells} should be named PGP1_21, PGP1_22, PGP1_A1
  • {ch} should be in the range 1-24
• hg38ToHg19 chainfile (hg38ToHg19.over.chain)
• samtools
• BCFtools
• Picard
• CrossMap
• HapCUT2
• Pypy
• Complete Genomics Variant Calls for PGP1 (GFF format)
• BAC library SNV calls for PGP1 (VCF format)
• Fragment boundaries for SISSOR libraries (called with HMM, see methods).