Bash scripts and codes are provided to process PacBio long-read RNA-seq data and generate isoform-level master table. The SLURM settings in all bash scripts are cluster-specific. Adjust memory, CPUs, job array size, and email before submitting on your system.

The pipeline executes the following steps in order:

1. align_pacbio.sh

   • Aligns PacBio reads to the reference genome using minimap2.

2. before_collapse.sh

   • Prepares GTFs and PSL files for downstream collapse steps.

3. collapse_pacbio.sh

   • Collapses isoforms using custom Python scripts to produce a non-redundant isoform set.

4. sqanti_IR.sh

   • Runs SQANTI3 for isoform QC and filters intron retention events.

5. cpat_nmd.sh

   • Runs CPAT to analyze coding potential and filters nonsense-mediated decay (NMD) transcripts.

6. generate_master_table.R

   • Generates a comprehensive master table of all isoforms and genes, including:
     • isoform categories (protein coding, NMD, noORF, RI)
     • canonical/annotated transcript annotation
     • entropy/perplexity
     • TPM expression levels
     • exon counts
     • splicing event integration
     • ORF-level summaries

Download environment.yaml file to figure out all dependencies and avoid manual installation of each package. To create a Conda environment from the file, use:

conda env create -f encode_pacbio.yml
conda activate encode_pacbio

Before running this script, you should confirm that all fastq files you wish to process are in a single directory. Then, create a textfile containing all filenames of fastq files you wish to process.

first.fastq.gz
second.fastq.gz
third.fastq.gz

You should also confirm that you have a file containing chromosome sizes for all chromosomes. You can generate one using pyfaidx as follows:

faidx /path/to/genome/reference.fa -i chromsizes > /path/to/output/chrNameLength.txt

In the bash script, insert correct user-defined variables for 1) a path to output directory 2) a path to genome reference 3) genome annotationg gtf 4) a chromosome length file 5) path to the directory where all raw fastq files are stored, and 6) textfile containing filenames that we craeted above.

For 6) path to FLAIR helper scripts, insert a path to where all FLAIR helper scripts are stored. You can download from FLAIR github directly, or just download files shared in this github repo.

In the bash script, insert correct user-defined variables for 1) path to input directory where all output files of align_pacbio.sh are stored in 2) genome annotation gtf 3) a path to ouput directory 4) a prefix for subset gtf files that will be generated from the input genome annotation gtf.

Also note that this script assumes the data are human data with chromosomes 1-22, X, and Y. Since this is hard-coded, if you wish to run this script on different species, please edit the numbers in the loop

for i in {1..22} X Y; do
    ...
done

Before running this script, you should generate a manifest.tsv file that contains relevant information for each aligned fastq file. This file should only contain three columns, being (1) sample accession (2) tissue information, and (3) path to fastq file.

ENCFF011DHZ prefrontal_cortex /path/to/ENCFF011DHZ.fastq.gz
ENCFF012NOH adrenal_gland /path/to/ENCFF012NOH.fastq.gz
...

You should also have two bed files that contain relevant information about acceptable TSS and TES range (one bed file each). In our paper, ...

bash/run_pacbio_pipeline.sh directs to all bash and R scripts needed to run the entire pipeline. For each file, edit user-defined variables to run on your data.