HoMi

HoMi is your friend for Host-Microbiome dual transcriptome data processing.

What is HoMi?

HoMi is a pipeline developed to streamline processing host-microbiome dual transcriptome data, but it can also work with solely metagenomic or metatranscriptomic data for host filtering and mapping-based processing to taxonomic and functional profiles.

HoMi manages software environments and deployment on a Slurm-managed compute cluster, but it can also be run locally. For statistics tools, see the R package HoMiStats.

HoMi is currently under development! Please feel free to raise an issue or contribute.

Installation

mamba create -n HoMi python=3.11
conda activate HoMi
pip install homi-pipeline

Usage

HoMi.py <config_file> --cores <n_cores> --profile <profile_name>

Config file

An example config file is provided in tests/example_config.yaml. This config file should contain paths to relevant files, like the metadata and databases. It can also be used to alter rule-specific resource requirements (details in example config).

Do you want to make sure your config file fits the requirements? Run check_config.py <config_filepath> to find out! (See src/check_config.py). HoMi will also automatically check your config before running the pipeline.

Metadata file

An example metadata file is provided in tests/example_metadata.csv. Metadata files should contain (at the minimum) a Sample column (named Sample), a forward reads filepath column (column name specified in the config file under fwd_reads_path), and a reverse reads filepath column (column name specified in the config file under rev_reads_path). These filepaths should be relative to the directory from which you are running HoMi.

Using HoMi on a cluster

If running HoMi on a cluster with SLURM, please setup a Snakemake SLURM profile. This will handle submitting batch jobs for each sample for each step of the pipeline. Then, pass this the name of this profile to HoMi.py <config_file> --profile <profile_name>, no cores need to be passed.

src/profile_setup.py is a script that can be used to setup a cluster profile for Slurm integration, with options for clusters with and without hyperthreaded cores.

Conda environment building

If conda environments have already been built, and you'd like snakemake to not build them, pass the argument --conda_prebuilt. This is particularly useful if running HoMi on a system with ARM architecture, like a Mac with M1/M2 chip.

Working directory

The working directory for HoMi can be set using the --workdir flag. This will set the "home base" for any relative filepaths within HoMi or your config, but it will not affect the filepath of the config or profile provided to HoMi.

Running with example dataset

# Create the mock community (too big for github)
python benchmarking/synthetic/create_mock_community.py tests/mock_community_sim_params.csv --work_dir tests/mock_community/

# Use the example config and example metadata provided
HoMi.py tests/example_config.yaml --cores 1

Unlocking a snakemake directory

Sometimes, when snakemake unexpectedly exits (e.g., due to a server connection timeout), the directory may be locked. Pass the argument --unlock to unlock the directory before running HoMi.py.

Pipeline steps

Preprocessing

1. Create a symbolic link to the sequencing files.
2. Trim reads using Trimmomatic
   • Trims adapter sequences from reads
   • Trims reads with a starting PHRED quality below readstart_qual_min
   • Trims reads with an ending PHRED quality below readend_qual_min
   • Trims reads with wherever there's a 4-base sliding window average PHRED quality score below 20
   • Removes reads with a length below min_readlen
3. Generates quality report with FastQC + MultiQC
4. Trim read ends using Seqtk
   • Consider this a second pass, in case Trimmomatic didn't catch something
5. Generates a second pass quality report with FastQC + MultiQC after the second trimming step

Read mapping

6. Remove host reads using Hostile

   • User should pass a database in the config file. Currently supported options are human-t2t-hla and human-t2t-hla-argos985.
   • OR users can pass a filepath to a bowtie2 index without the .bt extensions (e.g., index/example_index, where files exist named index/example_index.bt1, index/example_index.bt2, etc.).

7. Microbial read profiling

   • (A) Run HUMAnN pipeline on nonhost reads to profile microbial reads
     • Makes microshades taxa barplot from MetaPhlan and HUMAnN outputs
   • (B) Run Kraken + Bracken on nonhost reads to profile microbial taxa
     • Second taxonomy method for redundancy/validation/comparison

8. Align all reads against the host genome

   • HoMi will by default download the GRCh38 human reference genome, but you can provide an alternative genome (fna + gtf) if it's already downloaded
   • Either BBmap or HISAT2 is used to map the reads (HISAT2 by default), and featureCounts is used to generate a read count table

DAG

Using this pipeline for microbe-only samples

To use this pipeline for metagenomics/metatranscriptomics, you can add an (optional) column in the metadata, titled map_host. If map_host doesn't exist in your metadata, the entire pipeline (including mapping to the host genome) will be performed for all samples. If map_host exists, it should only contain boolean values (True/False), and the host genome will only be mapped for samples where map_host is True. This will still run host decontamination before microbial taxonomic/functional profiling.

Example:

Sample,map_host 
sample_nohost,False 
sample_withhost,True

Main repository contents

• snakefile contains the bulk of the pipeline
• src/ contains HoMi.py, a wrapper controlling the behavior of the snakemake pipeline, as well as other auxilliary utility scripts.
• conda_envs/ contains the conda environments for each rule in the snakemake pipeline
• data/ contains relevant data, such as adapter sequences to be removed during trimming.

Auxiliary files

• src/HoMi_cleanup.py contains a script that can be used to clean up unecessary intermediate files, if you decide you don't want them. Current functionality deletes temporary files from failed HUMAnN runs and HUMANnN .bam files across all samples provided in the metadata.
• src/profile_setup.py is a script that can be used to setup a cluster profile for Slurm integration, with options for clusters with and without hyperthreaded cores. Researchers at CU Boulder/Anschutz using Alpine should run this script as profile_setup.py --cluster_type slurm-nosmt to setup a Slurm snakemake profile compatible with Alpine, which does not have hyperthreaded cores.