Quick Run

## path to longcallR-dp, e.g. "/path/to/longcallR-dp"
cmd="/path/to/longcallR-dp"
## path to bam file, e.g. "path/to/aln.bam"
bam="demo/hg002_chr22.bam"
## path to reference genome, e.g. "/path/to/GRCh38.fa"
ref="demo/chr22.fa"
## output_path + prefix, e.g. "sample_longcallR_inference/contig"
out="hg002_chr22_feature/contig"
## minimum depth, default="6"
depth="6"
## minimum alternative allele fraction, default="0.1"
alt_frac="0.1"
## minimum base quality, hifi:0, ont:10, default="0"
min_bq="0"
## n_jobs, default="25"
n_jobs="1"
## threads, default="8"
threads="4"
## contigs, default="chr1 chr2 ... chr22, chrX chrY"
CTGS=("chr22")


## run longcallR-dp (Parallel)
parallel -j ${n_jobs} \
    "$cmd --mode predict \
    --bam-path '${bam}' \
    --ref-path '${ref}' \
    --min-depth '${depth}' \
    --min-alt-freq '${alt_frac}' \
    --min-baseq '${min_bq}' \
    --threads '${threads}' \
    --contigs {1} \
    --output '${out}_{1}'" ::: "${CTGS[@]}"

## download the configure files and models
longcallR_nn download --download_dir models

## run longcallR-nn (Serial)
for ctg in ${CTGS[@]}; do
    longcallR_nn call \
    -config models/pb_masseq_config.yaml \
    -model models/pb_masseq_model.chkpt \
    -data ${out}_${ctg} \
    -ref ${ref} \
    -output hg002_${ctg}.vcf \
    -max_depth 200 \
    -batch_size 500
done

Table of Contents

• Introduction
• Install
• Usage
• License

Introduction

<longcallR_nn> is a variant caller specifically designed for long-read RNA-seq data, utilizing a ResNet50 model. It provides high-precision identification of RNA SNPs and A-to-I RNA editing events. Currently, longcallR_nn is compatible with PacBio MAS-Seq/Iso-Seq and Nanopore cDNA/dRNA data.

Installation

Prerequisites

• Python>=3.9
• Rust
• Parallel

Dependencies

• Pytorch>=1.3

Install from source code

1. Clone the repo:

   git clone https://github.com/huangnengCSU/longcallR-nn.git

2. Navigate to the project directory:

   cd longcallR-nn

3. Install the longcallR-nn dependencies:

   # create env
   conda create -n longcallRenv python=3.9
   # activate env
   conda activate longcallRenv
   # install dependencies
   pip install -r requirements.txt
   # install PyTorch gpu version (pytorch>=1.3)
   conda install pytorch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 pytorch-cuda=12.4 torchmetrics -c pytorch -c nvidia
   # or install PyTorch cpu version (pytorch>=1.3)
   conda install pytorch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 torchmetrics cpuonly -c pytorch

4. Install longcallR-nn:

   pip install .

5. Compile longcallR-dp:

   cd longcallR_dp
   cargo build --release

Install from Bioconda

1. Install longcallR-nn:

   # create env
   conda create -n longcallRenv python=3.9
   # activate env
   conda activate longcallRenv
   
   # install PyTorch gpu version (pytorch>=1.3)
   conda install pytorch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 pytorch-cuda=12.4 -c pytorch -c nvidia
   # or install PyTorch cpu version (pytorch>=1.3)
   conda install pytorch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 cpuonly -c pytorch
   
   # install longcallR-nn from bioconda
   conda install -c bioconda -c conda-forge longcallr_nn

2. Compile longcallR-dp:

   git clone https://github.com/huangnengCSU/longcallR-nn.git
   cd longcallR-nn/longcallR-dp
   cargo build --release

Apptainer/Singularity

You can build an Apptainer/Singularity container using the provided longcallR-nn.def definition file with the following command:

cd longcallR-nn
apptainer build --fakeroot longcallR-nn.sif longcallR-nn.def

Alternatively, you can pull a pre-built image directly with:

apptainer pull docker://huangnengcsu/longcallr_nn:v0.0.1_gpu

Docker

To build a Docker image, run the following command:

cd longcallR-nn
docker build -t longcallr_nn:v0.0.1_gpu .

Alternatively, you can pull the Docker image directly with:

docker pull huangnengcsu/longcallr_nn:v0.0.1_gpu

Usage

longcallR-dp extract training or inference dataset

Template scripts for generating training dataset or inference dataset can be found in the longcallR_dp/job_templates directory. The script for generating a REDIportal VCF file for training A-to-I RNA-editing events is located at longcallR_dp/scripts/REDIportal_to_vcf.py.

Usage: longcallR_dp [OPTIONS] --bam-path <BAM_PATH> --ref-path <REF_PATH> --mode <MODE> --output <OUTPUT>

Options:
  -b, --bam-path <BAM_PATH>
          Path to input bam file
  -f, --ref-path <REF_PATH>
          Path to reference file
  -m, --mode <MODE>
          Mode: train or predict [possible values: train, predict]
      --truth <TRUTH>
          Truth vcf file path (Optional)
      --editing <EDITING>
          RNA editing dataset (Optional)
      --passed-editing <PASSED_EDITING>
          Passed RNA editing dataset by comparing with alignment (Optional)
      --bed-path <BED_PATH>
          Interested region
  -o, --output <OUTPUT>
          Output bam file path
  -r, --region <REGION>
          Region to realign (Optional). Format: chr:start-end, left-closed, right-open
  -x, --contigs [<CONTIGS>...]
          Contigs to be processed. Example: -x chr1 chr2 chr3
  -t, --threads <THREADS>
          Number of threads, default 1 [default: 1]
      --min-mapq <MIN_MAPQ>
          Minimum mapping quality for reads [default: 10]
      --min-baseq <MIN_BASEQ>
          Minimum base quality for allele [default: 0]
      --min-alt-freq <MIN_ALT_FREQ>
          Minimum allele frequency for candidate SNPs [default: 0.1]
      --min-alt-depth <MIN_ALT_DEPTH>
          Minimum alternate allele depth for candidate SNPs [default: 2]
      --min-depth <MIN_DEPTH>
          Minimum depth for candidate SNPs [default: 6]
      --max-depth <MAX_DEPTH>
          Maximum depth for candidate SNPs [default: 20000]
      --min-read-length <MIN_READ_LENGTH>
          Minimum read length to filter reads [default: 500]
      --flanking-size <FLANKING_SIZE>
          Flanking size [default: 20]
      --chunk-size <CHUNK_SIZE>
          Chunk size [default: 5000]
  -h, --help
          Print help
  -V, --version
          Print version

longcallR-nn call variants

Table 1: Configuration files and pretrained models

The configuration files and models for ONT and PacBio long-read RNA-seq have been released on Zenodo. You can use the command longcallR_nn download to automatically retrieve the latest versions of these files.

Platform	config	model
PacBio Masseq	hg002_na24385_masseq.yaml	hg002_na24385_mix_nopass_resnet50_sgd.epoch30.chkpt
PacBio Isoseq	hg002_isoseq.yaml	hg002_baylor_isoseq_nopass_resnet50_sgd.epoch30.chkpt
ONT cDNA	wtc11_cdna.yaml	cdna_wtc11_nopass_resnet50_sgd.epoch30.chkpt
ONT dRNA	hg002_drna.yaml	drna_hg002_nopass_resnet50_sgd2.epoch31.chkpt

Once you have obtained the model files, you can run the following command to call variants:

longcallR_nn call [-h] -config CONFIG -model MODEL -data DATA [-ref REF] -output OUTPUT [-max_depth MAX_DEPTH] [-batch_size BATCH_SIZE] [--no_cuda]

optional arguments:
  -h, --help                show this help message and exit
  -config CONFIG            path to config file
  -model MODEL              path to trained model
  -data DATA                directory of feature files
  -ref REF                  reference genome file
  -output OUTPUT            output vcf file
  -max_depth MAX_DEPTH      max depth threshold
  -batch_size BATCH_SIZE    batch size
  --no_cuda                 If running on cpu device, set the argument.

longcallR-nn train model

The template configuration file to refer to is config/test.yaml.

longcallR_nn train [-h] -config CONFIG [-log LOG]

optional arguments:
  -h, --help      show this help message and exit
  -config CONFIG  path to config file
  -log LOG        name of log file

License

MIT License

Copyright (c) 2024 Dana-Farber Cancer Institute.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.