Releases: ENCODE-DCC/chip-seq-pipeline2
Releases · ENCODE-DCC/chip-seq-pipeline2
v2.2.2
v2.2.1
- Fixed the issue that test run takes too much resources. Now users can run a test pipeline on a laptop.
- Fixed conflicts in Conda installer.
- Updated documentation:
- Pipeline still supports Conda as an optional method, DACC will no longer help fixing Conda conflicts on user's computer. If any error occurs, then switch to Singularity method.
- More comprehensive and simpler and README
v2.2.0
Upgrade Caper to >=2.2.0 for new HPC features.
$ pip install caper>=2.2.0 --upgrade
Conda environment name change:
encode-chip-seq-pipeline
->encd-chip
- Please reinstall your Conda environment after updating pipeline's git repo.
$ scripts/uninstall_conda_env.sh $ scripts/install_conda_env.sh
Caper's new HPC feature
- Unified command
caper hpc
for all HPC backend types (slurm
,sge
,pbs
andlsf
). - Read Caper's README carefully and make a backup of your Caper's conf file (
~/.caper/default.conf
) and runcaper init YOUR_BACKEND
. caper hpc submit chip.wdl -i INPUT.json --leader-job-name GOOD_NAME_FOR_LEADER [--conda | --singularity]
: submits a workflow to HPC's job engine.caper hpc list
: shows all Caper leader jobscaper hpc abort JOB_ID
: terminate a Caper leader job including child jobs
v2.1.6
Fixed python2 shebang
STDERR_BACKGROUND_CONTENTS=
Traceback (most recent call last):
File "/home/zhaosj/anaconda3/envs/encode-atac-seq-pipeline/bin/encode_task_filter.py", line 438, in <module>
main()
File "/home/zhaosj/anaconda3/envs/encode-atac-seq-pipeline/bin/encode_task_filter.py", line 344, in main
args.nth, args.mem_gb, args.out_dir)
File "/home/zhaosj/anaconda3/envs/encode-atac-seq-pipeline/bin/encode_task_filter.py", line 136, in rm_unmapped_lowq_reads_pe
res_param=get_samtools_res_param('fixmate', nth=nth),
File "/home/zhaosj/anaconda3/envs/encode-atac-seq-pipeline/bin/encode_lib_common.py", line 359, in run_shell_cmd
raise Exception(err_str)
Exception: PID=44241, PGID=44241, RC=127, DURATION_SEC=0.0
STDERR=/usr/bin/env: ‘python2’: No such file or directory
v2.1.5
v2.1.4
v2.1.2
v2.1.1
v2.1.0
Hotfix for chip.subsample_reads
bug.
This bug was introduced in v1.2.2 (6/13/2019).
chip.ctl_subsample_reads
is ignored and chip.subsample_reads
(0: disabled by default) is used for subsampling both experiment replicates and controls.
Outputs are affected by this bug if chip.subsample_reads
is non-zero and chip.pipeline_type
is tf
(default) or histone
.
Since chip.subsample_reads
is 0 (disabled) by default, outputs are not affected if you used a default set of parameters.
The following outputs can be affected by this bug:
- Control TAG-ALIGNs and all its downstream analyses.
- Such downstream outputs include all kinds of peaks (spp, macs2, idr, overlap).