Comments (16)
Hi,
You are rigth – a 9 bp UMI is quite short. As such, we're seeing about half the reads being dropped due to multiple CDR3s being assigned to the same UMI. Considering the UMIs are attached to multiple V gene primers, a good way around might be to include a few nucleotides right after the UMI, potentially increasing diversity. I'd recommend giving this a go: --tag-pattern "^(UMI:N{15})(R1:*)"
or maybe even longer to capture the difference between primers. If that's not cutting it, let me know, and we can tinker around with the parameters to try and save more reads. Although a non-UMI approach is also a good choice since MiXCR has very poverful error-correction algorithms even for data without barcodes.
As for the CDR3 discrepancy. In the paper they do include an extra amino acid from the FR4 (sourced from the J gene) within the CDR3. The reasoning behind this addition isn't entirely clear. While some researchers opt to exclude the initial and final amino acids from the CDR3 definition (e.i. IMGT), adding an extra one is a bit weird However, since this particular amino acid stems from the J gene – which both methods identify correctly – you can safely consider the clones equivalent.
For a quick comparison:
- CASSQDSDPQGLFAGNTIYFG (from the paper)
- CASSQDSDPQGLFAGNTIYF (MiXCR)
Check out this link, and you'll see that the terminal 'G' belongs to the FR4.
from mixcr.
Actually, the 15bp UMI looks much better; the number of unassigned alignments has dropped from 53% to 7.8%! Additionally, 85% of the reads are used in clonotype assembly.
I would suggest going with the 15bp UMI. While it's not perfect, it still allows you to leverage the UMI to correct the data effectively.
from mixcr.
HI, yes, of course, could you please share the size of the UMI?
from mixcr.
Size of the UMI is 7 bp long. it is pretty short.
from mixcr.
It's pretty short. I must say I have read this paper before and contacted the authors on the matter of sharing the data (because, as far as I'm concerned, the raw data is not publicly available), so I can tune the preset, but I never heard back from them. If you have raw data generated by this protocol that you can share with us (the same goes for the single-part data from this publication) - that would be of great help.
Nevertheless, here is the command I suggest using, judging from the scheme:
mixcr analyze generic-amplicon-with-umi \
--species hsa \
--rna \
--tag-pattern "^(R1:*)\(UMI:N{10})(R2:*)" \
--floating-left-alignment-boundary \
--floating-right-alignment-boundary C \
input_R1.fastq.gz \
input_R2.fastq.gz \
result
Because the UMI is quite short, I suggest trying to include a few more letters from the TRBC/TRAC primer, which will at least increase the diversity two-fold.
from mixcr.
Hello, I do have the bulk and single cell data generated from this protocol. I would probably have to converse with the data generator because the data that we have is a clinical data of origin. I will get back to you once I talked with the developer and get back to you.
in terms of the bulk data, would single pair of the Fastq file suffice? ( R1 and R2 ) Also for the single cell data, Would you need all fastq files for the entire batch(it will be 384 pairs of fastq files in total)?
from mixcr.
A single pair of files will be enough for our purposes. In the case of Single-cell analysis, it's better to see the full picture, as the filtering process includes all cells. If needed, we can provide a secure SFTP server for the data transfer.
Nevertheless, I recommend you try the commands suggested and we can see how well it worked, as these generic presets should cover most cases.
from mixcr.
the command above throws an error stating that, "Could not invoke public final void com.milaboratory.mixcr.cli.AlignMiXCRMixins.floatingRightAlignmentBoundary(java.lang.String) with /jsimonlab/users/bshim/BMS-Bulk-Reads/BMS-61_S1_L001_R1_001.fastq.gz (java.lang.IllegalArgumentException: Unknown point: /jsimonlab/users/bshim/BMS-Bulk-Reads/BMS-61_S1_L001_R1_001.fastq.gz)"
why might this be?
from mixcr.
Please try the following:
mixcr analyze generic-amplicon-with-umi \
--species hsa \
--rna \
--tag-pattern "^(R1:*)\(UMI:N{10})(R2:*)" \
--floating-left-alignment-boundary \
--floating-right-alignment-boundary C \
input_R1.fastq.gz \
input_R2.fastq.gz \
result
from mixcr.
I am also in the process of getting access to sample data for both single and bulk library which we can share to you. I will let you guys know as soon as possible.
from mixcr.
Upon analyzing the bulk dataset, I see that as expected, such a short UMI sequence leads to a high number of distinct clones within a single UMI group, which in some cases makes it hard to assemble consensus. I tweaked the parameters in the example below to recover as many clones as possible.
mixcr analyze generic-amplicon-with-umi \
--species hsa \
--rna \
--tag-pattern "^(R1:*)gaagcaga\^(UMI:N{11}) || ^(R1:*)taccagct\^(UMI:N{11})" \
--floating-left-alignment-boundary \
--floating-right-alignment-boundary C \
-Massemble.consensusAssemblerParameters.assembler.maxIterations=10 \
-Massemble.consensusAssemblerParameters.assembler.minRecordSharePerConsensus=0.01 \
-Massemble.consensusAssemblerParameters.assembler.minRecursiveRecordShare=0.01 \
-Massemble.consensusAssemblerParameters.assembler.maxConsensuses=10 \
input_R1_001.fastq.gz \
input_R2_001.fastq.gz \
output
Nevertheless, it is strongly recommended using a longer UMI, as in this case it doesn't really mark unique molecules, thus de facto is not a true UMI.
Alternatively, you can analyze the data ignoring the UMI sequence. In your case there is no need in R2 file at all then, as it doesn't cover anything but a portion of C gene.
mixcr analyze generic-amplicon \
--species hsa \
--rna \
--tag-pattern "^(R1:*)gaagcaga || ^(R1:*)taccagct" \
--floating-left-alignment-boundary \
--floating-right-alignment-boundary C \
input_R1_001.fastq.gz \
output
Sincerely,
Mark
from mixcr.
Could you explain little bit about the tag pattern used here?
are the 8bp sequences after ^(R1:*), index 1 and index 2 for every sample? What do the 8bp sequences exactly represent here?
Also, single cell presets are still in a working progress I am assuming?
from mixcr.
In your R1 files the reads have UMI and Illumina indices at the end. These 8bp is the small part of C gene at the very end of the payload sequence (that is most likely comes from the primer) that I use to trim artificial barcode sequences.
The single-cell preset is still work in progress, I will get back to you with it later this week.
from mixcr.
Hi @mizraelson,
I recently read one paper entitled TCR sequencing and cloning methods for repertoire analysis and isolation of tumor-reactive TCRs. In this paper, they introduced one TCR sequencing method for RNA extracted from T cells under the name SEQTR
. The library structure is [UMI 9 bases][VDJ][C constant region]
, and the sequencing strategy is SE150. I downloaded the raw sequencing files from GEO website and analyzed GSM7061297 (SRR23603384) with the following protocol:
# Step 1. Trim adaptor.
fastp -i SRR23603384_1.fastq.gz -o SRR23603384_trimmed_1.fastq.gz -w 8
# Step 2. Analyze the data with UMI assigned as the first 10 bases.
# From the supplementary file of the paper, I learned that the 9-base UMI is HHHHHNNNN,
# Then I calculated the presence of G in the first 9 bases of each trimmed fastq, it turned
# out that the first base had a higher frequency of G. So I chose to use the first 10 bases
# as UMI. Maybe I should have chosen 1 to 10 bases as UMI?
mixcr analyze generic-amplicon-with-umi \
--threads 16 \
--species hsa \
--rna \
--rigid-left-alignment-boundary \
--floating-right-alignment-boundary C \
--tag-pattern '^(UMI:N{10})(R1:*)' \
-Massemble.consensusAssemblerParameters.assembler.maxIterations=6 \
-Massemble.consensusAssemblerParameters.assembler.minRecordSharePerConsensus=0.02 \
-Massemble.consensusAssemblerParameters.assembler.minRecursiveRecordShare=0.1 \
-Massemble.consensusAssemblerParameters.assembler.maxConsensuses=6 \
../fastqs/SRR23603384_trimmed_1.fastq.gz \
output
# Alternative Step 2. Ignore UMI and run mixcr by trimming the first 10 bases.
# After read this post and several post discussing UMI, I think 9-base UMI is too short.
mixcr analyze generic-amplicon \
--threads 16 \
--species hsa \
--library imgt \
--rna \
--rigid-left-alignment-boundary \
--floating-right-alignment-boundary C \
--tag-pattern '^N{10}(R1:*)' \
../fastqs/SRR23603384_trimmed_1.fastq.gz \
noUMI
The qc output for Step 2
is:
Successfully aligned reads: 97.36% [OK]
Off target (non TCR/IG) reads: 0.27% [OK]
Reads with no V or J hits: 2.36% [OK]
Reads with no barcode: 0.0% [OK]
Alignments that do not cover CDR3: 0.48% [OK]
Tag groups that do not cover CDR3: 0.018% [OK]
Barcode collisions in clonotype assembly: 86.56% [ALERT]
Unassigned alignments in clonotype assembly: 53.29% [ALERT]
Reads used in clonotypes: 44.95% [ALERT]
Alignments dropped due to low sequence quality: 1.75% [OK]
Clones dropped in post-filtering: 0.0% [OK]
Alignments dropped in clones post-filtering: 0.0% [OK]
Reads dropped in tags error correction and filtering: 0.93% [OK]
UMIs artificial diversity eliminated: 12.31% [OK]
Reads dropped in UMI error correction and whitelist: 0.0% [OK]
Reads dropped in tags filtering: 0.93% [OK]
The qc output for Alternative step 2
is:
Successfully aligned reads: 97.36% [OK]
Off target (non TCR/IG) reads: 0.32% [OK]
Reads with no V or J hits: 2.31% [OK]
Reads used in clonotypes: 95.62% [OK]
Alignments that do not cover CDR3: 0.48% [OK]
Alignments dropped due to low sequence quality: 2.10% [OK]
Clones dropped in post-filtering: 0.0% [OK]
Alignments dropped in clones post-filtering: 0.0% [OK]
Then I compared the output files of TRB.tsv
with the results published by the authors. I found there is one amino acid difference for the most abundant clones.
For example, the first five line from the results published by the authors reads:
#CDR3_sequence Count TRBV TRBJ Frame CDR3_aaseq CDR3_length
TGCGCCAGCAGCCAAGATTCCGATCCCCAGGGGCTGTTTGCGGGAAACACCATATATTTTGGA 174591 hTRBV04-3 hTRBJ01-3 IN CASSQDSDPQGLFAGNTIYFG 21
TGTGCCAGCAGCCAAGGGACAGGACGGTCTTCACCCCTCCACTTTGGG 158629 hTRBV03-1 hTRBJ01-6 IN CASSQGTGRSSPLHFG 16
TGTGCCAGCTCACCGACAGGGGAGGCCACTGAAGCTTTCTTTGGA 127792 hTRBV18 hTRBJ01-1 IN CASSPTGEATEAFFG 15
TGCCAGCAGCTCTTAGCGCAATCCGTTCTTCGGG 87563 hTRBV21 hTRBJ02-1 OUT _ _
TGTGCCAGCAGTTTCCCGGATACGCAGTATTTTGGC 80302 hTRBV28 hTRBJ02-3 IN CASSFPDTQYFG 12
For the results from Step 2
, the first five line reads:
cloneId readCount readFraction uniqueMoleculeCount uniqueMoleculeFraction targetSequences targetQualities allVHitsWithScore allDHitsWithScore allJHitsWithScore allCHitsWithScore allVAlignments allDAlignments allJAlignments allCAlignments nSeqCDR3 minQualCDR3 aaSeqCDR3 refPoints
0 162907.0 0.031750372110455366 17152 0.027187463840134162 TGCGCCAGCAGCCAAGATTCCGATCCCCAGGGGCTGTTTGCGGGAAACACCATATATTTT [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV4-3*00(563.6) TRBD1*00(30) TRBJ1-3*00(458.2) TRBC1*00(50.5) 347|365|384|0|18||180.0 16|22|36|27|33||30.0 24|42|70|42|60||180.0 TGCGCCAGCAGCCAAGATTCCGATCCCCAGGGGCTGTTTGCGGGAAACACCATATATTTT 58 CASSQDSDPQGLFAGNTIYF :::::::::0:1:18:27:-4:-2:33:42:-4:60:::
3 82214.0 0.016023406561344676 8470 0.013425712379077446 TGTGCCAGCAGCCAAGGGACAGGACGGTCTTCACCCCTCCACTTT [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV3-1*00(521.2),TRBV3-2*00(520) TRBD1*00(35) TRBJ1-6*00(437.8) TRBC1*00(141.5) 347|363|384|0|16||160.0;347|363|384|0|16||160.0 13|20|36|16|23||35.0 29|45|73|29|45||160.0 TGTGCCAGCAGCCAAGGGACAGGACGGTCTTCACCCCTCCACTTT 58 CASSQGTGRSSPLHF :::::::::0:-1:16:16:-1:-4:23:29:-9:45:::
1 81189.0 0.01582363533350783 10114 0.016031600354426127 TGTGCCAGCAGTTACGGGACAGTCTCTGGAAACACCATATATTTT [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV6-5*00(243.3) TRBD1*00(35) TRBJ1-3*00(498.5) TRBC1*00(106.4) 347|362|384|0|15||150.0 12|19|36|15|22||35.0 20|42|70|23|45||220.0 TGTGCCAGCAGTTACGGGACAGTCTCTGGAAACACCATATATTTT 58 CASSYGTVSGNTIYF :::::::::0:-2:15:15:0:-5:22:23:0:45:::
2 65077.0 0.01268342653067151 8582 0.013603242460123097 TGTGCCAGCAGTTACGTTGGGGGTGGCTACACCTTC [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV6-5*00(242.5) TRBD1*00(25)TRBJ1-2*00(408.6) TRBC1*00(144.3) 347|362|384|0|15||150.0 18|23|36|18|23||25.0 27|40|68|23|36||130.0 TGTGCCAGCAGTTACGTTGGGGGTGGCTACACCTTC 58 CASSYVGGGYTF :::::::::0:-2:15:18:-6:-1:23:23:-7:36:::
6 49634.0 0.009673604997516015 4385 0.00695061969093915 TGTGCCAGCAGTGACTGGGGGGGGCAGGGAGCTTTCTTT [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV6-1*00(209.9) TRBD1*00(31)TRBJ1-1*00(378.3) TRBC1*00(117.1) 347|361|384|0|14||140.0 12|21|36|20|29|SA15G|31.0 30|40|68|29|39||100.0 TGTGCCAGCAGTGACTGGGGGGGGCAGGGAGCTTTCTTT 58 CASSDWGGQGAFF:::::::::0:-3:14:20:0:-3:29:29:-10:39:::
For the results from Alternative step 2
, the first five line reads:
cloneId readCount readFraction targetSequences targetQualities allVHitsWithScore allDHitsWithScore allJHitsWithScore allCHitsWithScore allVAlignments allDAlignments allJAlignments allCAlignments nSeqCDR3 minQualCDR3 aaSeqCDR3 refPoints
0 180598.0 0.01654797835251231 TGCGCCAGCAGCCAAGATTCCGATCCCCAGGGGCTGTTTGCGGGAAACACCATATATTTT [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV4-3*01(572.3) TRBD1*01(30) TRBJ1-3*01(459.8) TRBC1*01(50.5),TRBC1*02(50.5),TRBC1*03(50.5) 270|288|307|0|18||180.0 16|22|36|27|33||30.0 24|42|70|42|60||180.0 ;; TGCGCCAGCAGCCAAGATTCCGATCCCCAGGGGCTGTTTGCGGGAAACACCATATATTTT 58 CASSQDSDPQGLFAGNTIYF :::::::::0:1:18:27:-4:-2:33:42:-4:60:::
1 163591.0 0.014989647319825477 TGTGCCAGCAGCCAAGGGACAGGACGGTCTTCACCCCTCCACTTT [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV3-1*01(536.4),TRBV3-2*01(535.8) TRBD1*01(35) TRBJ1-6*02(438.3) TRBC1*01(142.1),TRBC1*02(142.1),TRBC1*03(142.1) 270|286|307|0|16||160.0;270|286|307|0|16||160.0 13|20|36|16|23||35.0 29|45|73|29|45||160.0 ;; TGTGCCAGCAGCCAAGGGACAGGACGGTCTTCACCCCTCCACTTT 58 CASSQGTGRSSPLHF :::::::::0:-1:16:16:-1:-4:23:29:-9:45:::
2 127701.0 0.011701089622222697 TGTGCCAGCTCACCGACAGGGGAGGCCACTGAAGCTTTCTTT [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV18*01(532) TRBD1*01(40) TRBJ1-1*01(438.9) TRBC1*01(153.4),TRBC1*02(153.4),TRBC1*03(153.4) 273|287|310|0|14||140.0 14|22|36|14|22||40.0 24|40|68|26|42||160.0 ;; TGTGCCAGCTCACCGACAGGGGAGGCCACTGAAGCTTTCTTT 58 CASSPTGEATEAFF :::::::::0:-3:14:14:-2:-2:22:26:-4:42:::
3 121440.0 0.011127401693978311 TGTGCCAGCAGTTACGGGACAGTCTCTGGAAACACCATATATTTT [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV6-5*01(185.2),TRBV6-2*01(183.6),TRBV6-3*01(183.6) TRBD1*01(35) TRBJ1-3*01(499) TRBC1*01(110.7),TRBC1*02(110.7),TRBC1*03(110.7) 270|285|307|0|15||150.0;270|285|307|0|15||150.0;270|285|307|0|15||150.0 12|19|36|15|22||35.020|42|70|23|45||220.0 ;; TGTGCCAGCAGTTACGGGACAGTCTCTGGAAACACCATATATTTT 58 CASSYGTVSGNTIYF :::::::::0:-2:15:15:0:-5:22:23:0:45:::
4 109079.0 0.009994778074583828 TGTGCCAGCAGTTACGTTGGGGGTGGCTACACCTTC [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[ TRBV6-5*01(197.9) TRBD1*01(25),TRBD2*01(25) TRBJ1-2*01(408.4) TRBC1*01(147.9),TRBC1*02(147.9),TRBC1*03(147.9) 270|285|307|0|15||150.0 18|23|36|18|23||25.0;25|30|48|18|23||25.0 27|40|68|23|36||130.0 ;; TGTGCCAGCAGTTACGTTGGGGGTGGCTACACCTTC 58 CASSYVGGGYTF :::::::::0:-2:15:18:-6:-1:23:23:-7:36:::
I truly value your expertise and insight in this matter and I believe your perspective could be of great help.
Best,
Jianxiang
from mixcr.
Hi,
You are rigth – a 9 bp UMI is quite short. As such, we're seeing about half the reads being dropped due to multiple CDR3s being assigned to the same UMI. Considering the UMIs are attached to multiple V gene primers, a good way around might be to include a few nucleotides right after the UMI, potentially increasing diversity. I'd recommend giving this a go:
--tag-pattern "^(UMI:N{15})(R1:*)"
or maybe even longer to capture the difference between primers. If that's not cutting it, let me know, and we can tinker around with the parameters to try and save more reads. Although a non-UMI approach is also a good choice since MiXCR has very poverful error-correction algorithms even for data without barcodes.As for the CDR3 discrepancy. In the paper they do include an extra amino acid from the FR4 (sourced from the J gene) within the CDR3. The reasoning behind this addition isn't entirely clear. While some researchers opt to exclude the initial and final amino acids from the CDR3 definition (e.i. IMGT), adding an extra one is a bit weird However, since this particular amino acid stems from the J gene – which both methods identify correctly – you can safely consider the clones equivalent.
For a quick comparison:
- CASSQDSDPQGLFAGNTIYFG (from the paper)
- CASSQDSDPQGLFAGNTIYF (MiXCR)
Check out this link, and you'll see that the terminal 'G' belongs to the FR4.
Thank you for your clarification of the "G" amino acid shown in the results of the manuscript.
I have both run the pipeline with set the first 15 bases as UMI and the first 25 bases as UMI. The results are slightly different. The results for the first 15 bases set as UMI is:
Successfully aligned reads: 97.51% [OK]
Off target (non TCR/IG) reads: 0.46% [OK]
Reads with no V or J hits: 2.021% [OK]
Reads with no barcode: 0.0% [OK]
Alignments that do not cover CDR3: 0.42% [OK]
Tag groups that do not cover CDR3: 0.32% [OK]
Barcode collisions in clonotype assembly: 69.56% [ALERT]
Unassigned alignments in clonotype assembly: 7.69% [WARN]
Reads used in clonotypes: 85.67% [WARN]
Alignments dropped due to low sequence quality: 6.13% [OK]
Clones dropped in post-filtering: 0.0% [OK]
Alignments dropped in clones post-filtering: 0.0% [OK]
Reads dropped in tags error correction and filtering: 4.44% [OK]
UMIs artificial diversity eliminated: 11.94% [OK]
Reads dropped in UMI error correction and whitelist: 0.0% [OK]
Reads dropped in tags filtering: 4.44% [OK]
The results for the first 25 bp as UMI is:
Successfully aligned reads: 97.16% [OK]
Off target (non TCR/IG) reads: 1.66% [OK]
Reads with no V or J hits: 1.17% [OK]
Reads with no barcode: 0.0% [OK]
Alignments that do not cover CDR3: 0.087% [OK]
Tag groups that do not cover CDR3: 0.041% [OK]
Barcode collisions in clonotype assembly: 63.57% [ALERT]
Unassigned alignments in clonotype assembly: 5.76% [WARN]
Reads used in clonotypes: 85.98% [WARN]
Alignments dropped due to low sequence quality: 7.88% [OK]
Clones dropped in post-filtering: 0.0% [OK]
Alignments dropped in clones post-filtering: 0.0% [OK]
Reads dropped in tags error correction and filtering: 5.87% [WARN]
UMIs artificial diversity eliminated: 12.21% [OK]
Reads dropped in UMI error correction and whitelist: 0.0% [OK]
Reads dropped in tags filtering: 5.87% [WARN]
When the first 10 bases are ignored using the fore-mentioned Alternative step 2
, the results shows:
Successfully aligned reads: 97.36% [OK]
Off target (non TCR/IG) reads: 0.32% [OK]
Reads with no V or J hits: 2.31% [OK]
Reads used in clonotypes: 95.62% [OK]
Alignments that do not cover CDR3: 0.48% [OK]
Alignments dropped due to low sequence quality: 2.10% [OK]
Clones dropped in post-filtering: 0.0% [OK]
Alignments dropped in clones post-filtering: 0.0% [OK]
Should I try to use longer bases to be used as UMI, or should I just ignore the first 10 bases?
Thank you very much!
Best,
Jianxiang
from mixcr.
Thank you for your clarification!
Best,
Jianxiang
from mixcr.
Related Issues (20)
- Analysis of single chain Fragment variable (scFv) HOT 1
- Error exportPlots shmTrees - Duplicate library: repseqio.v4.0_with_found_alleles:10090 HOT 2
- transitioning from postanalysis to exportPlots HOT 1
- Somatic Hypermutation Status of IGHV status HOT 2
- Question about filtering reads prior to Mixcr HOT 1
- Preset for long-read RNAseq with cell barcode HOT 1
- How to perform MiXCR on my spatial TCR data HOT 1
- The problem of mixcr exportPlots shmTrees HOT 3
- The majority of results in 'clones_TR[A|B].tsv' are "region_not_covered" HOT 1
- allVHitsWithScore allDHitsWithScore allJHitsWithScore allCHitsWithScore HOT 1
- FASTQ and BAM give discordant results HOT 6
- Help with TCR-seq alignment rate
- findAlleles get empty clone data HOT 11
- Rhapsody BCR+TCR full length data input HOT 2
- Generating all all_contig_annotations.json to run enclone after MiXCR
- Postanalysis output explanation HOT 3
- `Feature for allele search doesn't intersect JGene` error when running `findAlleles` HOT 1
- Runtime error of "Can't apply step BuildingInitialTrees" when running findShmTrees HOT 3
- Question: Does mixcr take sequence stagger into consideration? HOT 1
- ERROR: picocli.CommandLine$ExecutionException: Error while running command align java.lang.IllegalStateException: Check failed. HOT 6
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from mixcr.