hku-bal / clair3-trio Goto Github PK

Where can I download the training data chr1 you used in Figure 5?Thank you very much！

Figure 5. Comparison of different architectures and shapes for calling variants from trios. All results were trained on chr1 and tested on chr20.

Hi,
I am trying to run a Clair3-Trio analysis on a trio using the script run_clair3_trio.sh with the following options:
un_clair3_trio.sh
--bam_fn_c=child.bam
--bam_fn_p1=parent1.bam
--bam_fn_p2=parent2.bam
--output=output_dir/
--ref_fn=ref.fasta
--threads=12
--model_path_clair3=/path/to/Clair3-Trio/models/clair3_models/r941_prom_hac_g360+g422
--model_path_clair3_trio=/path/to/Clair3-Trio/models/clair3_trio_models/c3t_hg002_g422
--whatshap=/path/to/clair3_env/bin/whatshap
--include_all_ctgs
--enable_output_phasing
--enable_output_haplotagging

where the *bam files have been produced aligning the raw reads against my ref genome using minimap2.
However, it seems that the runtime exceeds 50 hours (which is my walltime limit). In addition, no output file have been produced.

Is there anyone who had the same problems? Is there any option I have to change in order to speed up the analysis or something I am actually missing?

Thank you for your help!

Hi, I am toying with an idea to collect sample for a family with a condition of interest and there is 3 generation available, grandparent do not have conditon while 1 parent and 2 child has the condition. Can this kind of family structure be used effectively to identify variant? Also I currently only identified 1 family so far, does it work with 1 family or more family is required?

I downloaded GM24385_1_Guppy_4.2.2_prom.fastq.gz according to Training Data guppy4.2.2 -- -- GM24385_5_Guppy_4.2.2_prom.fastq.gz, The coverage of BAM file obtained after merge and minimap is 72.79, but I see that the coverage of HG002 in training_data.md file is 432.38. I want to ask in the chain of https://s3-us-west-2.amazonaws.com/human-pangenomics/index.html?prefix=NHGRI_UCSC_panel/HG002/nanopore/Guppy_4.2.2/ ，Which fq files should I download?

Dear author：
In the creating tensors for model training part, i see
each tensors are generated from combination of trio samples
e.g. in the following example, we generation different combination from [10x, 30x, 60x, 80x] data from HG002 trio into
combination of (Child, Parent1, Parent2): (10x, 10x, 10x), (30x, 10x, 10x)

I trained with this example and got similar results to the paper on 10X data sets, but did not work well on 20X and above data sets. I would like to ask you what method you used in your final training to generate tensor.
Thank you！

Clair-trio won't output gVCF format when enabling "--gvcf"

There is no printed error other than:

cat: 'test/tmp/gvcf_tmp_output/*.tmp.g.vcf': No such file or directory
FileNotFoundError: [Errno 2] No such file or directory: 'SortVcf'

Let me know if there is any other info you need. Thanks.

Hi,

I would like to know whether there is any plan to upload more trio models like Clair3 (rerio) in near future? Also, are the pre-trained trio models trained with Ashkenazim family samples (HG002-4) only?

Thank you.
Jimmy

Hi,
Really thank you for the tool. Are there any plans to add also a clair3-Trio-illumina for Illumina short reads and a pre-trained model to use clair3-Trio with Illumina long reads?

Hello! I want to obtain Q20+ data (FASTQs and BAMs format) for experimental purposes. However, I found on the main page that almost all of them need AWS to download, is there any other way or website? Since I am a student without a VISA card, I cannot register an AWS account.
Hope to get your help

Hi, I am running Clair3-Trio on HTCondor and I am experiencing issues while "merging variants and non-variants to GVCF" for each chromosome, prompting both MemoryError and OSError: [Errno 12] Cannot allocate memory
It just pass to the next chromosome, prompting the same error and go on, until the last chromosome and then pass to the next tasks. At the end of the process, .vcfs and .bam files are correctly created (in size of GBs), meanwhile .gvcf are basically empty (size of few Mbs)

Apart from the huge walltime I did not run into issues while running the same script without the --gvcf flag, with these condor settings:
request_cpus = 16
request_memory = 128 GB

I encounter this issue with --gvcf even boosting the settings to:
request_cpus = 64
request_memory = 512 GB
So I reckon it's not a matter of lack of memory from my side.

This is an example of the error prompted:

[INFO] Pileup variants processed in chr1: 12927629
[INFO] Trio variants processed in chr1: 1722491
[INFO] Merge variants and non-variants to GVCF
Traceback (most recent call last):
File "/home/Clair3-Trio/trio/../clair3.py", line 112, in
main()
File "/home/Clair3-Trio/trio/../clair3.py", line 106, in main
submodule.main()
File "/home/Clair3-Trio/trio/MergeVcf_Trio.py", line 259, in main
mergeNonVariant(args)
File "/home/Clair3-Trio/trio/MergeVcf_Trio.py", line 180, in mergeNonVariant
args.ctgEnd)
File "/home/Clair3-Trio/preprocess/utils.py", line 221, in mergeCalls
self.writeNonVarBlock(curNonVarStart, curNonVarEnd, curNonVarPos, curNonVarCall, save_writer)
File "/home/Clair3-Trio/preprocess/utils.py", line 181, in writeNonVarBlock
self._writeRightBlock(start, end, curNonVarCall, save_writer)
File "/home/Clair3-Trio/preprocess/utils.py", line 164, in _writeRightBlock
new_ref = self.readReferenceBaseAtPos(pos_cmd)
File "/home/Clair3-Trio/preprocess/utils.py", line 152, in readReferenceBaseAtPos
reader = os.popen(cmd)
File "/home/miniconda3/envs/clair3/lib-python/3/os.py", line 980, in popen
bufsize=buffering)
File "/home/miniconda3/envs/clair3/lib-python/3/subprocess.py", line 744, in init
restore_signals, start_new_session)
File "/home/miniconda3/envs/clair3/lib-python/3/subprocess.py", line 1323, in _execute_child
restore_signals, start_new_session, preexec_fn)
OSError: [Errno 12] Cannot allocate memory
[INFO] Pileup variants processed in chr4: 15631830
...

And this is the script I am running, I emptied the arguments by purpose:

home/Clair3-Trio/run_clair3_trio.sh \
  --bam_fn_c=/home/new_hg38/minimap_S2_sorted.bam \
  --bam_fn_p1=/home/new_hg38/minimap_S3_sorted.bam \
  --bam_fn_p2=home/new_hg38/minimap_S4_sorted.bam \
  --output=/home/new_trio_variants \
  --ref_fn=/home/new_hg38/hg38.fa \
  --threads=12 \
  --model_path_clair3=/home/Clair3-Trio/models/clair3_models/r941_prom_hac_g360+g422 \
  --model_path_clair3_trio=/home/Clair3-Trio/models/clair3_trio_models/c3t_hg002_g422 \
  --whatshap=/home/miniconda3/envs/clair3/bin/whatshap \
  --gvcf \
  --include_all_ctgs \
  --enable_output_phasing \
  --enable_output_haplotagging

Do someone experienced the same issue?

When using --enable_phasing I can't see any phased SNV on my final VCF. This happens in v0.4 and v0.5.

Hi there,

could this be useful for non-human trio data?

Many thanks!
CW