Contact: Ruibang Luo, Junhao Su
Email: rbluo@cs.hku.hk, jhsu@cs.hku.hk
Clair3-Trio is archived. Please visit the new version Clair3-Nova that extended Clair3-Trio to support accurate de novo variant calling.
Accurate identification of genetic variants from family child-mother-father trio sequencing data is important in genomics. However, state-of-the-art approaches treat variant calling from trios as three independent tasks, which limits their calling accuracy for Nanopore long-read sequencing data. For better trio variant calling, we introduce Clair3-Trio, the first variant caller tailored for family trio data from Nanopore long-reads. Clair3-Trio employs a Trio-to-Trio deep neural network model, which allows it to input the trio sequencing information and output all of the trio’s predicted variants within a single model to improve variant calling. We also present MCVLoss, a novel loss function tailor-made for variant calling in trios, leveraging the explicit encoding of the Mendelian inheritance. Clair3-Trio showed comprehensive improvement in experiments. It predicted far fewer Mendelian inheritance violation variations than current state-of-the-art methods.
Detailed descriptions of the methodology and results for Clair3-Trio are available in this paper.
- Introduction
- Latest Updates
- What's New in Clair3-Trio
- Pre-trained Models
- Quick Demo
- Installation
- Output Files
- Usage
- Folder Structure and Submodule Descriptions
- Clair3-Trio Model Training
- Training Data
- Analysis Results
- VCF/GVCF Output Formats
- Publication
v0.7.1 (Oct 17, 2023): Fix memory error issue.
v0.7 (July 9, 2023): Added the source/version/command tag into the VCF header. Fixed a bug of AF for 1/2 genotypes. Added AD into VCF output. Added document for output files. Added a page for the method of merging VCF.
v0.6 (April 25, 2023): Bumped up Python from 3.6 to 3.9, Whatshap from v1.0 to v1.7 Clair3 #193. Fixed gVCF format mistake #3. Added options "--enable_phasing", "--enable_output_phasing", and "enable_output_haplotagging" #4. Added singularity support.
v0.5 (April 10, 2023): Added support for gVCF output. Use --gvcf to enable gVCF output.
v0.4 (March 22, 2023): Added a model for R10.4 pore with the Kit 14 chemistry (Q20+). Check this page for more information about the model.
v0.3 (June 20, 2022): Optimized Clair3-Trio speed, the runtime of Clair3-Trio to call variants from the whole genome is about 2.4 times of calling a single sample form Clair3 (v0.1-r10).
v0.2 (May 15, 2022): A guppy5 model for Clair3-Trio is available now. Check this page for more information about the Guppy5 model.
v0.1 (April 22, 2022): Initial release.
- New Architecture. Clair3-Trio employs a Trio-to-Trio deep neural network model that allows it to take three samples as input and output the varaints of all three samples in one go.
- Mendelian violations aware. Clair3-Trio uses MCVLoss to improve variants calling in trio by penalizing mendelian violoations.
- Improved Performance. Using only 10x of HG002, 3 and 4 ONT data, Clair3-Trio achieved 97.30% SNP F1-score and 56.48% Indel F1-score. Compared to Clair3, Clair3-Trio reduced SNP errors by ~78%, and Indel errors by ~22%. Clair3-Trio signficantly reduced Mendelian violations from 48,345 to 7,072.
Download models from here or click on the links below.
| Model name | Platform | Training samples | Date | Basecaller | File | Link |
|---|---|---|---|---|---|---|
| c3t_hg002_dna_r1041_e82_400bps_sup | ONT 10.4.1 | HG002,3,4 | 20230322 | Dorado v4.0.0 SUP | c3t_hg002_dna_r1041_e82_400bps_sup.tar.gz | Download |
| c3t_hg002_r941_prom_sup_g5014 | ONT r9.4.1 | HG002,3,4 | 20220514 | Guppy5 sup | c3t_hg002_r941_prom_sup_g5014.tar.gz | Download |
| c3t_hg002_g422 | ONT r9.4.1 | HG002,3,4 | 20220422 | Guppy4 hac | c3t_hg002_g422.tar.gz | Download |
When using the Clair3-Trio model, please use a corresponding Clair3 model for Pileup calling. Check here or here for more information about Clair3 pretrained model.
| Model name | Platform | Training samples | Date | Basecaller | File | Link |
|---|---|---|---|---|---|---|
| r1041_e82_400bps_sup_v400 | ONT r10.4.1 | HG002,4,5 | - | Dorado v4.0.0 SUP | r1041_e82_400bps_sup_v400.tar.gz | Download |
| r941_prom_sup_g5014 | ONT r9.4.1 | HG002,4,5 (Guppy5_sup) | 20220112 | Guppy5 sup | r941_prom_sup_g5014.tar.gz | Download |
| r941_prom_hac_g360+g422 | ONT r9.4.1 | HG001,2,4,5 | 20210517 | Guppy3,4 hac | r941_prom_hac_g360+g422.tar.gz | Download |
- see Trio Quick Demo.
A pre-built docker image is available here. With it you can run Clair3-Trio using a single command.
Caution: Absolute path is needed for both INPUT_DIR and OUTPUT_DIR.
INPUT_DIR="[YOUR_INPUT_FOLDER]" # e.g. /input
REF=${_INPUT_DIR}/ref.fa # change your reference file name here
OUTPUT_DIR="[YOUR_OUTPUT_FOLDER]" # e.g. /output
THREADS="[MAXIMUM_THREADS]" # e.g. 8
MODEL_C3="[Clair3 MODEL NAME]" # e.g. Clair3 model, r941_prom_hac_g360+g422 for Guppy4 data, r941_prom_sup_g5014 for Guppy5 data
MODEL_C3T="[Clair3-Trio MODEL NAME]" # e.g. Clair3-Trio model, c3t_hg002_g422 for Guppy4 data, c3t_hg002_r941_prom_sup_g5014 for Guppy5 data
docker run -it \
-v ${INPUT_DIR}:${INPUT_DIR} \
-v ${OUTPUT_DIR}:${OUTPUT_DIR} \
hkubal/clair3-trio:latest \
/opt/bin/run_clair3_trio.sh \
--ref_fn=${INPUT_DIR}/ref.fa \ ## change your reference file name here
--bam_fn_c=${INPUT_DIR}/child_input.bam \ ## change your child's bam file name here
--bam_fn_p1=${INPUT_DIR}/parent1_input.bam \ ## change your parent-1's bam file name here
--bam_fn_p2=${INPUT_DIR}/parenet2_input.bam \ ## change your parent-2's bam file name here
--sample_name_c=${SAMPLE_C} \ ## change your child's name here
--sample_name_p1=${SAMPLE_P1} \ ## change your parent-1's name here
--sample_name_p2=${SAMPLE_P2} \ ## change your parent-2's name here
--threads=${THREADS} \ ## maximum threads to be used
--model_path_clair3="/opt/models/clair3_models/${MODEL_C3}" \
--model_path_clair3_trio="/opt/models/clair3_trio_models/${MODEL_C3T}" \
--output=${OUTPUT_DIR} ## absolute output path prefix
Caution: Absolute path is needed for both INPUT_DIR and OUTPUT_DIR.
INPUT_DIR="[YOUR_INPUT_FOLDER]" # e.g. /input
REF=${_INPUT_DIR}/ref.fa # change your reference file name here
OUTPUT_DIR="[YOUR_OUTPUT_FOLDER]" # e.g. /output
THREADS="[MAXIMUM_THREADS]" # e.g. 8
MODEL_C3="[Clair3 MODEL NAME]" # e.g. Clair3 model, r941_prom_hac_g360+g422 for Guppy4 data, r941_prom_sup_g5014 for Guppy5 data
MODEL_C3T="[Clair3-Trio MODEL NAME]" # e.g. Clair3-Trio model, c3t_hg002_g422 for Guppy4 data, c3t_hg002_r941_prom_sup_g5014 for Guppy5 data
conda config --add channels defaults
conda create -n singularity-env -c conda-forge singularity -y
conda activate singularity-env
# singularity pull docker pre-built image
singularity pull docker://hkubal/clair3-trio:latest
singularity exec \
-B ${INPUT_DIR},${OUTPUT_DIR} \
clair3-trio_latest.sif \
/opt/bin/run_clair3_trio.sh \
--ref_fn=${INPUT_DIR}/ref.fa \ ## change your reference file name here
--bam_fn_c=${INPUT_DIR}/child_input.bam \ ## change your child's bam file name here
--bam_fn_p1=${INPUT_DIR}/parent1_input.bam \ ## change your parent-1's bam file name here
--bam_fn_p2=${INPUT_DIR}/parenet2_input.bam \ ## change your parent-2's bam file name here
--sample_name_c=${SAMPLE_C} \ ## change your child's name here
--sample_name_p1=${SAMPLE_P1} \ ## change your parent-1's name here
--sample_name_p2=${SAMPLE_P2} \ ## change your parent-2's name here
--threads=${THREADS} \ ## maximum threads to be used
--model_path_clair3="/opt/models/clair3_models/${MODEL_C3}" \
--model_path_clair3_trio="/opt/models/clair3_trio_models/${MODEL_C3T}" \
--output=${OUTPUT_DIR} ## absolute output path prefix
Anaconda install:
Please install anaconda using the official guide or using the commands below:
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
chmod +x ./Miniconda3-latest-Linux-x86_64.sh
./Miniconda3-latest-Linux-x86_64.shInstall Clair3 env and Clair3-Trio using anaconda step by step:
# create and activate an environment named clair3
conda create -n clair3 python=3.9.0 -y
source activate clair3
# install pypy and packages in the environemnt
conda install -c conda-forge pypy3.6 -y
pypy3 -m ensurepip
pypy3 -m pip install mpmath==1.2.1
# install python packages in environment
conda install -c conda-forge tensorflow==2.8.0 -y
conda install -c conda-forge pytables -y
conda install -c anaconda pigz cffi==1.14.4 -y
conda install -c conda-forge parallel=20191122 zstd -y
conda install -c conda-forge -c bioconda samtools=1.15.1 -y
conda install -c conda-forge -c bioconda whatshap=1.7 -y
conda install -c conda-forge xz zlib bzip2 automake curl -y
# tensorflow-addons is required in training
pip install tensorflow-addons
# clone Clair3-Trio
git clone https://github.com/HKU-BAL/Clair3-Trio.git
cd Clair3-Trio
# download Clair3's pre-trained models
mkdir -p models/clair3_models
wget http://www.bio8.cs.hku.hk/clair3_trio/clair3_models/clair3_models.tar.gz
tar -zxvf clair3_models.tar.gz -C ./models/clair3_models
# download Clair3-Trio's pre-trained models
mkdir -p models/clair3_trio_models
wget http://www.bio8.cs.hku.hk/clair3_trio/clair3_trio_models/clair3_trio_models.tar.gz
tar -zxvf clair3_trio_models.tar.gz -C ./models/clair3_trio_models
# run clair3-trio
_INPUT_DIR="[YOUR_INPUT_FOLDER]" # e.g. ./input
_BAM_C=${_INPUT_DIR}/input_child.bam # chnage your child's bam file name here
_BAM_P1=${_INPUT_DIR}/input_parent1.bam # chnage your parent-1's bam file name here
_BAM_P2=${_INPUT_DIR}/input_parent2.bam # chnage your parent-2's bam file name here
_SAMPLE_C="[Child sample ID]" # child sample ID, e.g. HG002
_SAMPLE_P1="[Parent1 sample ID]" # parent1 sample ID, e.g. HG003
_SAMPLE_P2="[Parent2 sample ID]" # parent2 sample ID, e.g. HG004
_REF=${_INPUT_DIR}/ref.fa # change your reference file name here
_OUTPUT_DIR="[YOUR_OUTPUT_FOLDER]" # e.g. ./output
_THREADS="[MAXIMUM_THREADS]" # e.g. 8
_MODEL_DIR_C3="[Clair3 MODEL NAME]" # e.g. ./models/clair3_models/ont
_MODEL_DIR_C3T="[Clair3-Trio MODEL NAME]" # e.g. ./models/clair3_trio_models/c3t_hg002_g422
./run_clair3_trio.sh \
--bam_fn_c=${_BAM_C} \
--bam_fn_p1=${_BAM_P1} \
--bam_fn_p2=${_BAM_P2} \
--output=${_OUTPUT_DIR} \
--ref_fn=${_REF} \
--threads=${_THREADS} \
--model_path_clair3="${_MODEL_DIR_C3}" \
--model_path_clair3_trio="${_MODEL_DIR_C3T}" \
--sample_name_c=${_SAMPLE_C} \
--sample_name_p1=${_SAMPLE_P1} \
--sample_name_p2=${_SAMPLE_P2}
# make sure channels are added in conda
conda config --add channels defaults
conda config --add channels bioconda
conda config --add channels conda-forge
# create conda environment named "clair3-trio"
conda create -n clair3-trio -c bioconda clair3-trio python=3.9.0 -y
conda activate clair3-trio
# run clair3-trio like this afterward
_INPUT_DIR="[YOUR_INPUT_FOLDER]" # e.g. ./input
_BAM_C=${_INPUT_DIR}/input_child.bam # chnage your child's bam file name here
_BAM_P1=${_INPUT_DIR}/input_parent1.bam # chnage your parent-1's bam file name here
_BAM_P2=${_INPUT_DIR}/input_parent2.bam # chnage your parent-2's bam file name here
_SAMPLE_C="[Child sample ID]" # child sample ID, e.g. HG002
_SAMPLE_P1="[Parent1 sample ID]" # parent1 sample ID, e.g. HG003
_SAMPLE_P2="[Parent2 sample ID]" # parent2 sample ID, e.g. HG004
_REF=${_INPUT_DIR}/ref.fa # change your reference file name here
_OUTPUT_DIR="[YOUR_OUTPUT_FOLDER]" # e.g. ./output
_THREADS="[MAXIMUM_THREADS]" # e.g. 8
_MODEL_DIR_C3="[Clair3 MODEL NAME]" # e.g. r941_prom_sup_g5014
_MODEL_DIR_C3T="[Clair3-Trio MODEL NAME]" # e.g. c3t_hg002_r941_prom_sup_g5014
run_clair3_trio.sh \
--bam_fn_c=${_BAM_C} \
--bam_fn_p1=${_BAM_P1} \
--bam_fn_p2=${_BAM_P2} \
--output=${_OUTPUT_DIR} \
--ref_fn=${_REF} \
--threads=${_THREADS} \
--model_path_clair3="${CONDA_PREFIX}/bin/models/${_MODEL_DIR_C3}" \
--model_path_clair3_trio="${CONDA_PREFIX}/bin/models/${_MODEL_DIR_C3T}" \
--sample_name_c=${_SAMPLE_C} \
--sample_name_p1=${_SAMPLE_P1} \
--sample_name_p2=${_SAMPLE_P2}Check Usage for more options. Pre-trained models are already included in the bioconda package.
Building a docker image.
# clone Clair3-Trio
git clone https://github.com/hku-bal/Clair3-Trio.git
cd Clair3-Trio
# build a docker image named hkubal/clair3-trio:latest
# might require docker authentication to build docker image
docker build -f ./Dockerfile -t hkubal/clair3-trio:latest .
# run clair3-trio docker image like
docker run -it hkubal/clair3-trio:latest /opt/bin/run_clair3_trio.sh --help
Clair3-Trio outputs files in VCF/GVCF format for the trio genotype. The output files (for a trio [C ], [P1], [P2]) including:
.
├── run_clair3_trio.log # Clair3-Trio running log
├── [C ].vcf.gz # Called variants in vcf format for [C ]
├── [P1].vcf.gz # Called variants in vcf format for [P1]
├── [P2].vcf.gz # Called variants in vcf format for [P2]
├── [C ].gvcf.gz # Called variants in gvcf format for [C ] (when enabled `--gvcf`)
├── [P1].gvcf.gz # Called variants in gvcf format for [P2] (when enabled `--gvcf`)
├── [P2].gvcf.gz # Called variants in gvcf format for [P2] (when enabled `--gvcf`)
├── phased_[C ].vcf.gz # Called phased variants for [C ] (when enabled `--enable_output_phasing`)
├── phased_[P1].vcf.gz # Called phased variants for [P1] (when enabled `--enable_output_phasing`)
├── phased_[P2].vcf.gz # Called phased variants for [P2] (when enabled `--enable_output_phasing`)
├── phased_[C ].bam # alignment tagged with phased variants info. for [C ] (when enabled `--enable_output_haplotagging`)
├── phased_[P1].bam # alignment tagged with phased variants info. for [P1] (when enabled `--enable_output_haplotagging`)
├── phased_[P2].bam # alignment tagged with phased variants info. for [P2] (when enabled `--enable_output_haplotagging`)
├── [C ]_c3t.vcf.gz # raw variants from Clair-Trio's trio model for [C ]
├── [P1]_c3t.vcf.gz # raw variants from Clair-Trio's trio model for [P1]
├── [P2]_c3t.vcf.gz # raw variants from Clair-Trio's trio model for [P2]
├── /log # folder for detailed running log
└── /tmp # folder for all running temporary files
Caution: Use =value for optional parameters, e.g. --bed_fn=fn.bed instead of --bed_fn fn.bed.
./run_clair3_trio.sh \
--bam_fn_c=${_BAM_C} \
--bam_fn_p1=${_BAM_P1} \
--bam_fn_p2=${_BAM_P2} \
--output=${_OUTPUT_DIR} \
--ref_fn=${_REF} \
--threads=${_THREADS} \
--model_path_clair3="${_MODEL_DIR_C3}" \
--model_path_clair3_trio="${_MODEL_DIR_C3T}" \
--bed_fn=${_INPUT_DIR}/quick_demo.bed \
--sample_name_c=${_SAMPLE_C} \
--sample_name_p1=${_SAMPLE_P1} \
--sample_name_p2=${_SAMPLE_P2}
Required parameters:
--bam_fn_c=FILE Child's BAM file input. The input file must be samtools indexed.
--bam_fn_p1=FILE Parent1's BAM file input (Parent1 can be father or mother). The input file must be samtools indexed.
--bam_fn_p2=FILE Parent2's BAM file input (Parent2 can be father or mother). The input file must be samtools indexed.
-f, --ref_fn=FILE FASTA reference file input. The input file must be samtools indexed.
--model_path_clair3=STR The folder path containing a Clair3 model (requiring six files in the folder, including pileup.data-00000-of-00002, pileup.data-00001-of-00002 pileup.index, full_alignment.data-00000-of-00002, full_alignment.data-00001-of-00002 and full_alignment.index.
--model_path_clair3_trio=STR The folder path containing a Clair3-Trio model.
-t, --threads=INT Max threads to be used. The full genome will be divided into small chunks for parallel processing. Each chunk will use 4 threads. The chunks being processed simultaneously is ceil($threads/4)*3. 3 is the overloading factor.
-o, --output=PATH VCF output directory.Other parameters:
Caution: Use =value for optional parameters, e.g., --bed_fn=fn.bed instead of --bed_fn fn.bed
--sample_name_c=STR Define the sample name for Child to be shown in the VCF file.[Child].
--sample_name_p1=STR Define the sample name for Parent1 to be shown in the VCF file.[Parent1].
--sample_name_p2=STR Define the sample name for Parent2 to be shown in the VCF file.[Parent2].
--bed_fn=FILE Call variants only in the provided bed regions.
--vcf_fn=FILE Candidate sites VCF file input, variants will only be called at the sites in the VCF file if provided.
--ctg_name=STR The name of the sequence to be processed.
--qual=INT If set, variants with >$qual will be marked PASS, or LowQual otherwise.
--samtools=STR Path of samtools, samtools version >= 1.10 is required.
--python=STR Path of python, python3 >= 3.6 is required.
--pypy=STR Path of pypy3, pypy3 >= 3.6 is required.
--parallel=STR Path of parallel, parallel >= 20191122 is required.
--whatshap=STR Path of whatshap, whatshap >= 1.0 is required.
--chunk_size=INT The size of each chuck for parallel processing, default: 5Mbp.
--gvcf Enable GVCF output, default: disable.
--print_ref_calls Show reference calls (0/0) in vcf file, default: disable.
--include_all_ctgs Call variants on all contigs, otherwise call in chr{1..22,X,Y} and {1..22,X,Y}, default: disable.
--snp_min_af=FLOAT Minimum SNP AF required for a candidate variant. Lowering the value might increase a bit of sensitivity in trade of speed and accuracy, default: ont:0.08.
--indel_min_af=FLOAT Minimum INDEL AF required for a candidate variant. Lowering the value might increase a bit of sensitivity in trade of speed and accuracy, default: ont:0.15.
--pileup_model_prefix=STR EXPERIMENTAL: Model prefix in pileup calling, including $prefix.data-00000-of-00002, $prefix.data-00001-of-00002 $prefix.index. default: pileup.
--fa_model_prefix=STR EXPERIMENTAL: Model prefix in full-alignment calling, including $prefix.data-00000-of-00002, $prefix.data-00001-of-00002 $prefix.index, default: full_alignment.
--trio_model_prefix=STR EXPERIMENTAL: Model prefix in trio calling, including $prefix.data-00000-of-00002, $prefix.data-00001-of-00002 $prefix.index, default: trio.
--var_pct_full=FLOAT EXPERIMENTAL: Specify an expected percentage of low quality 0/1 and 1/1 variants called in the pileup mode for full-alignment mode calling, default: 0.3.
--ref_pct_full=FLOAT EXPERIMENTAL: Specify an expected percentage of low quality 0/0 variants called in the pileup mode for full-alignment mode calling, default: 0.3 for ilmn and hifi, 0.1 for ont.
--var_pct_phasing=FLOAT EXPERIMENTAL: Specify an expected percentage of high quality 0/1 variants used in Clair3 WhatsHap phasing, default: 0.8 for ont guppy5 and 0.7 for other platforms.
--enable_output_phasing Output phased variants using whatshap, default: disable.
--enable_output_haplotagging Output enable_output_haplotagging BAM variants using whatshap, default: disable.
--enable_phasing It means `--enable_output_phasing`. The option is retained for backward compatibility.Clair3-Trio shares the same folder structure as Clair3, except for an additional folder trio. For descriptions for Clair3 folder, please check Clair3's Descriptions for more inforamtion.
Submodules in clair3/ are for variant calling and model training. Submodules in preprocess are for data preparation.
For all the submodules listed below, you can use -h or --help for available options.
trio/ |
submodules under this folder are pypy incompatible, please run using python |
|---|---|
CheckEnvs_Trio |
Check the environment and validity of the input variables, preprocess the BED input if necessary, --chunk_size sets the chuck size to be processed per parallel job. |
SelectCandidates_Trio |
Select trio candidates for clair3-trio calling. |
CallVarBam_Trio |
Call variants using a trained model and three BAM files. |
SortVcf_Trio |
Sort Trio's VCF file. |
MergeTenorsBam_Trio |
Create and merge three tensors into trio's tensors. |
CallVariants_Trio |
Call variants using a trained model and merged tensors of candidate variants. |
model |
define Clair3-Trio model |
Training |
- |
SelectHetSnp_Trio |
Select heterozygous SNP candidates from pileup model and true set. |
Merge_Tenors_Trio |
Merge three tensors into trio's tensors. |
Tensor2Bin_Trio |
Convert trio's tensors into Bin file for training. |
Train_Trio |
Training a trio model using the RectifiedAdam optimizer. We also use the Lookahead optimizer to adjust the RectifiedAdam parameters dynamically. The initial learning rate is 1e-3 with 0.1 learning rate warm-up. Input a binary containing tensors created by Tensor2Bin_Trio. |
Evaluation |
- |
Check_de_novo |
Benchmark calling results in terms of de novo variants. |
Clair3-Trio trained its trio models using four GIAB samples (HG002, HG003 and HG004). All models were trained with chr20 excluded (including only chr1-19, 21, 22). All data links can be found at this page.
| Platform | Reference | Aligner | Training samples |
|---|---|---|---|
| ONT | GRCh38_no_alt | minimap2 | HG002,3,4 |
Clair3-Trio supports both VCF and GVCF output formats. Clair3-Trio uses VCF version 4.2 specifications. Specifically, Clair3-Trio adds a P INFO tag to the results called using a pileup model, and a T INFO tag to the results called using a trio model.
Clair3-Trio outputs a GATK-compatible GVCF format that passes GATK's ValidateVariants module. Different from DeepVariant that uses <*> to represent any possible alternative allele, Clair3-Trio uses <NON_REF>, the same as GATK.
Clair3-Trio GVCF files can be merged with GLNexus. A GLNexus caller-based configuration file is available Download.
Note that the reference call in VCF is called via Model, and Refcall in GVCF may be inferred from allele depth.
We left some comments for merging multiple VCF/GVCF here.
- Zheng Z, Li S, Su J, Leung AW, Lam TW, Luo R. Symphonizing pileup and full-alignment for deep learning-based long-read variant calling. Nature Computational Science. 2022 Dec;2(12):797-803.
- Su J, Zheng Z, Ahmed SS, Lam TW, Luo R. Clair3-trio: high-performance Nanopore long-read variant calling in family trios with trio-to-trio deep neural networks. Briefings in Bioinformatics. 2022 Sep;23(5):bbac301.