De novo assembly

First set up Zorn/Bascet according to the install instructions. This tutorial assumes that you have debarcoded the reads.

When to perform de novo assembly

De novo assembly is the process of reconstructing a genome (or transcriptome) from sequencing reads without using a reference genome as a guide. It is used primarily in microbiology, as research on eukaryotes is commonly done on organisms with known genomes. Possible exceptions are the assembly of cancer genomes, but assembly of large genomes is difficult, and Zorn/Bascet is currently optimized for the handling of smaller microbial genomes.

De novo assembly using SKESA

Assembly of each cell’s genome can be performed using SKESA:

(SLURM-compatible step)

### Assemble all genomes
BascetMapCellSKESA(
  bascetRoot,
  inputName = "filtered",
  outputName = "contigs"
)

This produce contigs for each cell, all grouped together in “contigs”. Note that both Skesa and Spades only assembles cells for which there are sufficient number of reads.

Investigating assemblies

The output is zip-files with contigs for each cell, separately. You can thus easily extract the contigs of interest and analyze with any tool of choice (not just Zorn/Bascet).

You can obtain the contigs for one cell like this (this is the fastest way):

b <- OpenBascet(bascetRoot, "contigs", bascetInstance)
contigs <- BascetReadFile(b, cellID = "<cell_id>", filename = "contigs.fa", as = "text")
CloseBascet(b)

We also provide convencience functions for getting a subset. The worked example below will work for future file formats as well, not just ZIP:

bascetRoot <- "/path/to/your/bascet"
bascetInstance <- GetDefaultBascetInstance()

# Get cell names
cells <- BascetCellNames(bascetRoot, "contigs", bascetInstance = bascetInstance)

# Pick the cells you want — e.g. all of them, or a subset
listCells <- cells$cell[1:10]

# Make sure the output dir exists
outputDir <- "contigs_out"
dir.create(outputDir, showWarnings = FALSE)

# Dump contigs — one <cellid>.fa file per cell
BascetDumpContigs(
  bascetRoot     = bascetRoot,
  inputName      = "contigs",          # shard holding contigs.fa per cell
  listCells      = listCells,
  outputDir      = outputDir,
  bascetInstance = bascetInstance
)

Further analysis of assembled contigs

To assess assemblies and annotate the contigs (QUAST, Bakta, Abricate, AMRfinder, GECCO, etc.), see the genome annotation vignette. For an overview of the underlying MAP framework, see Map scripts.

Coassembly

If you have cells with highly similar genomes, you might be able to generate “coassemblies” - consensus assemblies using reads from multiple similar cells.

To do this, simply extract the debarcoded reads from the cells of interest. There are many ways of picking them, where one method is to use the clustering function in Seurat. But you can use any method you wish to come up with a list of cell names. To learn more about clustering, see this Seurat tutorial first. Note that it is an open research problem how to best pick cells for coassembly!

After clustering, you can get the names of cells in cluster “0”:

### Get cell names
listCells <- rownames(adata[adata$cluster_id="0",])

Next extract a FASTQ with reads:

BascetDumpContigs(
  bascetRoot     = bascetRoot,
  inputName      = "contigs",          # shard holding contigs.fa per cell
  listCells      = listCells,
  outputDir      = outputDir,
  bascetInstance = bascetInstance
)

And finally, run SKESA or your favourite software to assemble the contigs: (in BASH)

#Run SKESA


########################################################### TODO: run skesa