Version 3.2

Added features

GitHub Actions testing scheme

1. Download and subset test datasets from ENCODE. Test data is ATAC-Seq of adrenal glands of people aged 59 and 16, 2 replicates each condition. Subset the FASTQs to 4 M reads per read per biological replicate.
2. Git clone the repo
3. Install miniconda and custom snakemake env (5.32) with plotly
4. Activate conda environment and gather basic intel
5. Snakemake dry run + create pipeline environments
6. Activate bwa env and make index of hg38 chr 1.
7. Run SnakeMake.

Version 3.1

Version 3.1

Added features

1. new consensus peak script. Take peaks that appear in n replicates per condition, then merge these peaks together.
2. scripts/make_differential_config.py helps to generate DESeq2 and DiffBind config files.
3. homer submits commands to SLURM instead of a local run.
4. preseq rules will now assess library complexity after all preprocessing steps. if lcextrap fails, the following restart attempts will be run in defect mode with a '-D' flag.
5. add functioning log files to deseq2 and diffbind.
6. new scheme to make bigwig files. Average CPM signal using a 10 bp window all with bedtools and chrom size file. DeepTools bamCoverage usually takes between 1 - 2 hrs or more, this method takes 20 - 30 mins.

Modifications

1. re-route output of fastp as the input for fastqc
2. fragment length files are tab-separated and the length is capped at 1 kbp.
3. combined mito removal + quality mapped reads rule into the filter rule.
4. reformatted all rules into Snakefile

Deprecated features

1. removed read shifting because we don't need base-pair resolution analyses (e.g. footprinting).
2. removed intervene rule.
3. removed tracking BAM entries before and after preprocessing steps.

need to do:

1. sensible github actions testing.

Version 3.0

Major code refactoring

Quality Control

QC results summarized into multiqc report.
QC include fastqc, fastq_screen, fragment length, FRiP.
No preseq. Needs extra rule after markdup to sort + index.

Processes

pipe bwa mem to samtools sort
Blacklist removal is abstracted to its own rule originally from consensus peak calling.
Rules are written in order from top to bottom.

New features

MultiQC report summarizes your results.
Fastp is preferred read trimmer.
HOMER integration! DESeq2 exports contrast combinations table, which is parsed and fed to HOMER I/O. Only runs if a contrast gets >= 10 DE up/down peaks.

Version 2.0

Version 2

• No more down-sampling of data.
• No more playing with regex to determine input files. Just symlink and go.
• Moderate code formatting
• essential_report.html aggregates important metrics in alignment, peak calling, and differential peak stages.

Version 1

• Choose to down-sample or not down-sample reads before analysis
• Made the format of the analysis: read trimming, alignment, remove mito + low-qual + dup reads, shift reads, take consensus peaks, and DESeq2.