`extract` requires three inputs: (1) a BAM alignment file of long-reads containing the ts:A flag (provided by default by Minimap2), (2) a reference genome and (3) a sorted and bgzip'd gene annotation file. Currently only gtf files are supported. We suggest using the `--junc-bed` parameter in minimap2 for more accurate spliced alignments. Without the parameter, it may be harder to distinguish unannoted exitron splicing events from annotated splicing events. To sort your gtf file, use the command: ```console $ awk '$1 ~ /^#/ {print $0;next} {print $0 | "sort -k1,1 -k4,4n -k5,5n"}' in.gtf > out_sorted.gtf ``` To bgzip your gene annotation file, use: ```console $ bgzip in.gtf ``` `bgzip` is part of the [htslib](http://www.htslib.org/), which you most likely already have installed if you care about BAM files. Otherwise, you can get it [here](http://www.htslib.org/). It is important to note that if you download the [latest GENCODE release](https://www.gencodegenes.org/human/) it will be in the gzip form, not bgzip. You will need to run `gzip -d` and then `bgzip`. ScanExitronLR utilizes the `gffutils` package, which requires an SQL-lite database of the annotation file. You do not need to provide such a file, as ScanExitronLR will create one if one is not found, though it may take ~20 minutes to build. It will be saved as `your_annotation.gtf.gz.db` in the same location as your annotation and will not need to be built again. In addition, we require a tabix index, and it will be created if one is not found. This should only take seconds. It will be saved as `your_annotation.gtf.gz.tbi`. Thus, if you are running ScanExitronLR on a shared server and using a shared annotation database, you may not have writing privelages in the shared space. You will need to copy your annotation file to your local directory. We have provided fully processed GTF files for [Gencode V39](https://drive.google.com/drive/folders/1LAU26BxAmTmkQdCaHJ1ba1LHBmOI8yUM?usp=sharing) and [TAIR10](https://drive.google.com/drive/folders/1FNZ5HRJOvGeiMxMObXBPgTGC2E0l3yeE?usp=sharing) for your convenience. To run ScanExitronLR in extract mode, simply run ```console $ selr extract ... ``` with the following parameters: | Parameter | Description | | --- | --- | | -i STR | REQUIRED: Input BAM file | | -g STR | REQUIRED: Input genome reference (e.g. hg38.fa) | | -r STR | REQUIRED: Input _sorted_ and bgzip'd annotation reference (e.g. gencode_v38_sorted.gtf.gz). | | -o STR | Output filename (e.g. bam_filename.exitron <- this is default) | | -a/--ao INT | Reports only exitrons with AO (# of supporting reads) of INT or above (default: 2). | | -p/--pso FLOAT | Reports only exitrons with PSO of FLOAT or above (default: 0.01). | | -c/--cores INT | Use INT cores (default: 1). Use as many as you can spare. Even large BAM files only use at most 4GB total memory on 10 cores. | | -cp/--cluster-purity FLOAT | Reports only exitrons with cluster purity of FLOAT or above (default: 0). | | -m/--mapq INT | Only considers reads with mapq score >= INT (default: 50) | | -j/--jitter INT | Treat splice-sites with fuzzy boundry of +/- INT (default: 10). | | -sr | Use this flag to skip the realignment step. | | -sa | Use this flag to save isoform abundance files for downstream differential isoform usage analysis with LIQA. Files are of the form: input.isoform.exitrons, input.isoform.normals (See example page) | We provide some functionality to perform exitron differential isoform usage with the `-sa` flag. See [here](https://github.com/ylab-hi/ScanExitronLR/tree/main/test_data) for an example.

ScanExitronLR filters exitron splicing events based on AO (-a/--ao), PSO (-p/--pso) and cluster purity (-cp/--cluster-purity): > __AO__. By default, ScanExitronLR only reports exitrons with at least two supporting reads (AO >= 2). This is filter out random sequencing errors that may lead to a faulty alignment and false splicing event. However, if the coverage is particularly low, you may need to set the AO threshold to 1 in order to detect exitrons in medium and lower expressed genes. > __PSO__. By default, ScanExitronLR only reports exitrons with a splicing frequence (PSO) above 1%. Splicing events below this frequency may not be biologically relevant or may just be due to splicing noise. Setting PSO filtering to 0% is not recommended because it will increase running time and report many low quality splicing events. > __Cluster Purity__. By default, ScanExitronLR does not filter by cluster purity. However, cluster purity is important for having high confidence the the reported splice sites. For example, if the cluster purity is 90%, then 90% of the exitron spliced reads have the reported splice sites. Thus, one ought to be cautious when investigating exitrons with cluster purities below 50%. There is an exitron splicing event being detected, but it is unclear where the exact splice sites occur. This can happen if the reads are particularly noisy or are aligned to a repetitive region.