# install the external bins with conda
conda install -c bioconda samtools
conda install -c bioconda bedtools
conda install -c bioconda minimap2

# use pip from pypi
pip install trackcluster
# or pip from source code for the latest version
git clone https://github.com/Runsheng/trackcluster.git
pip install ./trackcluster

# test if all dependencies are installed
trackrun.py test --install

# prepare the reference annotation bed file from gff file
# tested on Ensembl, WormBase and Arapost gff
gff2bigg.py -i ensemblxxxx.gff3 -o ref.bed 
# WormBase full gff contains too many information, need to extract the lines from WormBase only
cat c_elegans.PRJNA13758.WS266.annotations.gff3 |grep WormBase > ws266.gff
gff2bigg.py -i ws266.gff -o ref.bed
# the ref.bed can be sorted to speed up the analysis
bedtools sort -i ref.bed > refs.bed # refs.bed contains the sorted, know transcripts from gff annotation

# generate the read track from minimap2 bam file
bam2bigg.py -b group1.bam -o group1.bed
bam2bigg.py -b group2.bam -o group2.bed

# merge the bed file and sort
cat group1.bed group2.bed > read.bed
bedtools sort -i read.bed > reads.bed

# Examples for running commands:
trackrun.py clusterj -s reads.bed -r refs.bed -t 40 # run in junction mode, will generate the isoform.bed
trackrun.py count -s reads.bed -r refs.bed -i isoform.bed # generate the csv file for isoform expression
# alternative for cluster
trackrun.py cluster -s reads.bed -r refs.bed -t 40 # run in exon/intron intersection mode， slower, will generate the isoform.bed

# the post analysis could include the classification of novel isoforms
trackrun.py desc --isoform isoform.bed --reference ref.bed # generate the description for each novel isoform
# this part can be run directly on reads, to count the frequency of splicing events in reads, like intron_retention
trackrun.py addgene -r ref.bed -s reads.bed # will generate reads_gene.bed
trackrun.py desc --isoform reads_gene.bed --reference ref.bed # will generated reads_desc.txt and reads_class12.txt