Mosaic CLI

Basic Usage

Typing mosaic at the command line will yield a prompt like this:

$ mosaic

mosaic-*.*.*

Usage:
  mosaic bam [options] <sample-id> <BAM-or-CRAM>
  mosaic vcf [options] <VCF>
  mosaic gather [options] <dir>...
  mosaic init [options] <project-name>
  mosaic post [options] <project-id> <gathered-mosaic-json>
  mosaic map-paths <gathered-mosaic-json> <map-file></map-file>
  mosaic experiment [options] <project-id> <experiment-name> 

Run each command without any arguments (e.g. "mosaic bam") to see what [options] are available.

The philosophy of the CLI is to be easily usable with Unix-style tools and shell scripts, and to not attempt to parallelize or thread anything on its own, so as to easily fit in a variety of compute environments (e.g. batch, slurm, multicore EC2, etc.).

A simple example with 1 bam and 1 vcf looks like the following:

$ mosaic bam -d /data/bams NA21129 http://s3.amazonaws.com/1000genomes/phase3/data/NA21129/alignment/NA21129.chrom11.ILLUMINA.bwa.GIH.low_coverage.20121211.bam

$ mosaic vcf -d /data/vcfs -s NA21129 http://s3.amazonaws.com/frameshift1000g/all.vcf.gz (note: this VCF is large, so running will take a while. if you just want to test, I suggest just pulling out a subset first or using a smaller VCF)

$ mosaic gather /data/bams /data/vcfs

The above yields a file gathered.mosaic.json.gz, which is what we will POST to Mosaic.

Next, we set connection parameters for posting to a running Mosaic instance. $ export MOSAIC_URL=https://0.0.0.0:3000/ $ export [email protected] $ export MOSAIC_PASSWORD='my$password'

Alternatively, when using mosaic post you can set MOSAIC_TOKEN instead of MOSAIC_USERNAME and MOSAIC_PASSWORD. The token is an auth token that is provided by Mosaic. $ export MOSAIC_URL=https://0.0.0.0:3000/ $ export MOSAIC_TOKEN=be0ab7114beb70c14832a0c8de8ed708f2c1a5b8

Now, we create a new project on Mosaic (this step is optional; the project can already exist). The init command returns a project id for us to use if the creation was successful.

$ mosaic init my_project
project_id 2

Now, we post the data to this project: $ mosaic post 2 gathered.mosaic.json.gz

And we're done!

Using a config file

Instead of using environment variables, some commands offer the option of using the -c or --config option. With this option you will pass a filename of a config file instead of using environment variables. For example, you could have a file config.txt that looks like the following:

MOSAIC_URL=https://0.0.0.0:3000/
[email protected]
MOSAIC_PASSWORD=my$password

Then when calling MosaicCLI you use the following options:

$ mosaic init -c config.txt my_project
project_id 2

Details / Common issues

• You'll need to set LD_LIBRARY_PATH=<your-htslib-1.10-path>. htslib should be built with ./configure --enable-libcurl if you need to access AWS files.
• If using BED files, check that the chromosome names match. For instance, if the BED file has chr1 but your input files have 1, the CLI will error out.
• The -d flag passes an output directory.
• The use of http://s3.amazonaws.com/ instead of s3:// tells the tool to not try to use private credentials, and call the aws tool with --no-sign-request. This is an unfortunate artifact of the fact that samtools, bcftools, and aws have different conventions for how they handle public S3 files. In the future, this might be accomplished with a CLI flag instead.
• -s NA21129 tells mosaic vcf to only extract the one sample from the VCF file.
• Make sure to single-quote your password to prevent the shell from expanding any special characters.

More advanced usage

A more real-world use case with many samples and files might involve writing fuller bash scripts. The following example was used to produce the BAM output for 1000G files on a 96-core machine, the paths to which were provided in an index file 1000G.index2.

bucket="http://s3.amazonaws.com/1000genomes/phase3/"

# get just the bam column
tail -n +2 1000G.index2 | awk '{ print $1 }' > bam_files.txt

# do it in batches of 96.
ncores=96
while read line; do
      ((i=i%ncores)); ((i++==0)) && wait

      fpath="$bucket$line"
      sample_id=$(echo $line | cut -d '/' -f 2)

      mosaic bam -d /data/bam $sample_id $fpath &
done < bam_files.txt

rm bam_files.txt

S3 Files

In addition to local files, Mosaic CLI works with files stored on S3. If the S3 bucket is public and does not require credentials to access, you can format the URLs as http://s3.amazonaws.com/\/<file-name> and then use the CLI as if you were working with local files. If the bucket is private, you must set the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY variables in the environment or config file in order for Mosaic CLI to access the files. In order for Mosaic to be able to access the files, you must include the -i flag when running mosaic post. This will attach the S3 credentials in the request that gets sent to Mosaic.

Example:

export MOSAIC_URL=https://0.0.0.0:3000/
export MOSAIC_TOKEN=authtoken123
export AWS_ACCESS_KEY_ID=test123
export AWS_SECRET_ACCESS_KEY=test123
mosaic init test_project
mosaic bam -r 37 sample_id s3://bucket/file.bam
mosaic gather .
mosaic post -i project_id gathered.mosaic.json.gz

S3 compatible object storage

In addition to normal S3 files, Mosaic CLI works with S3 compatible object storage files. For Mosaic CLI to function correctly with these files, you will need to pass additional options to mosaic bam, mosaic vcf, and mosaic post. For mosaic bam and mosaic vcf you will need to pass the -e and -p params which are the S3 endpoint URL and S3 profile respectively while for mosaic post just the -e param needs to be passed. In order to use these options correctly, the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY variables must be set in the environement or in the config file and an additional file, ~/.s3cfg, must also exist. The .s3cfg file should be formatted as shown below:

[test_profile]
access_key = ****
secret_key = ****
host_base = https://test.com/

In this case, the -e option would be https://test.com/ and the -p option would be test_profile

Example:

export MOSAIC_URL=https://0.0.0.0:3000/
export MOSAIC_TOKEN=authtoken123
export AWS_ACCESS_KEY_ID=test123
export AWS_SECRET_ACCESS_KEY=test123
mosaic init test_project
mosaic bam -r 37 -e https://test.com/ -p test_profile sample_id https://test.com/file.bam
mosaic gather .
mosaic post -e https://test.com/ -i project_id gathered.mosaic.json.gz

Experiments

The CLI allows users to create experiments for a project by using the mosiac experiment command. This will return an experiment's ID which you can then use in mosaic bam and mosaic vcf with the -x parameter. An example is below

EXP_ID=$(mosaic experiment 1 test-experiment | cut -d ' ' -f 2)
mosaic bam -x $EXP_ID sample_id bam_url

Dependencies

You need at least samtools and bcftools executables, which the CLI calls out to. You will also need htslib version 1.10 which is used by mosdepth.

If you're dealing with S3 files, you need the aws executable as well.

Building from source (for dev team)

1. Build our local fork of mosdepth. Move the binary to bin/mosdepth.
2. Run go generate. This will cause fileb0x to produce a static folder. If this step does not work, you might need to update Go. Try a version >= go1.13.
3. Run go build. This will produce a mosaic executable.