passionlab / bella Goto Github PK

Hello,

I'm attempting to use BELLA but have encountered a problem in the alignment step.
The installation is one CentOS 3.10, using gcc 6.3.0, with virtualenv pip install of simplesam. When I try the test files using ./bella -i test.txt -o bella_output -d 30 the produced output file 'bella_output.out' is empty.

Looking over the print out I find this 'Average length of successful alignment -nan bp'. Everything else seems fine.

paeruginosa30x_0001_5reads.fastq: 0 MB
K-mer counting: BELLA
Output filename: bella_output.out
K-mer length: 17
X-drop: 7
Depth: 30X
Compute alignment: true
Seeding: two-kmer

Running with up to 4 threads
Reading FASTQ file paeruginosa30x_0001_5reads.fastq
Initial parsing, error estimation, and k-mer loading took: 0.0749692s

Cardinality estimate is 27251
Table size is: 186907 bits, 0.0222811 MB
Optimal number of hash functions is: 5
First pass of k-mer counting took: 0.0391129s
Second pass of k-mer counting took: 0.0153589s

Entries within reliable range: 231
Error rate estimate is -nan
Reliable lower bound: 2
Reliable upper bound: 30
Deviation from expected alignment score: 0.2
Constant of adaptive threshold: -nan

Running with up to 4 threads
Reading FASTQ file paeruginosa30x_0001_5reads.fastq
Fastq(s) parsing fastq took: 0.0317728s
Total number of reads: 5

Old number of nonzeros before merging: 474
New number of nonzeros after merging: 474
Old number of nonzeros before merging: 474
New number of nonzeros after merging: 474
Sparse matrix construction took: 0.00283812s

Available RAM is assumed to be: 8000 MB
FLOPS is 255
nnz(output): 10 | free memory: 8.38861e+09 | required memory: 288
Stages: 1 | max nnz per stage: 291271111

Columns [0 - 5] overlap time: 0.00734909s
Creating or appending to output file with 0 MB

Columns [0 - 5] alignment time: 0.0329758s | alignment rate: 937294 bases/s | average read length: 5556.6 | read pairs aligned this stage: 10
Average length of successful alignment -nan bps
Average length of failed alignment 5788.6 bps

Outputted 0 lines in 0.0115177s
Total running time: 0.346414s```

SELF_TODO: implement lower/upper bound as input parameter.

I'm getting the same error as issue #32 with a copy of the repository cloned today, in the same system background as described in issue #32 - Ubuntu 20.04, nvcc 11.6, gcc 9.4. The comment in that issue that LOGAN may have issues with "large-ish" input is a concern, because I'd like to use bella-gpu on a file containing about 600 gigabases of long-read sequences.
Two questions: (1) Is there a fix for the compile failure other than to install different versions of nvcc and gcc?
(2) If I'm successful in compiling, will bella-gpu process 600 gigabases of input?
Thanks!

I was doing some testing on my computer and I couldn't type the letter 'b' in the control centre

gcc -O3 -fopenmp -c -o bound.o kmercode/bound.cpp
In file included from /usr/include/c++/4.9/array:35:0,
from kmercode/bound.cpp:11:
/usr/include/c++/4.9/bits/c++0x_warning.h:32:2: error: #error This file requires compiler and library support for the ISO C++ 2011 standard. This support is currently experimental, and must be enabled with the -std=c++11 or -std=gnu++11 compiler options.
#error This file requires compiler and library support for the
^
make: *** [bound.o] Error 1

how can I solve the errors? thanks!

Hi,

I've been trying to use bella b150701 to assemble a small portion of the human genome. Whenever I try to use it though I run into the issue that bella apparently sometimes wants insane amounts of memory.

./bella -k 13 -i input.fastq -o test-bella -d 3

sometimes gives

@id1 : 133599645 MB
ACGT .... : 133599645 MB
: 133599645 MB

When I let this run, the memory fills up until I have to restart my computer.
Only sometimes i get somethig more feasible (with the exact same command, mind you) :

@id1 : 3 MB
ACGT .... : 3 MB
: 3 MB

I've tried this with different depth settings and with different k-mer lengths, but the problem seems to always occur.

Hi Giulia,

I've run into a problem when I try to run Bella using minimizers. Specifically, when I run the command

"bella -f input.txt -o output --paf -k 31 -w 17 -e 0.005 -l 2 -u 100",

I get the following output log:

INFO: src/../include/kmercount.hpp(108) InputFile = metagenome-anonymous.fastq
INFO: src/../include/kmercount.hpp(109) InputSize = 67.693069 MB
INFO: src/main.cpp(200) OutputFile = output.out
INFO: src/main.cpp(203) kmerSize = 31
INFO: src/main.cpp(206) GPUs = DISABLED
INFO: src/main.cpp(209) UserDefinedMemory = 8000.000000 MB
INFO: src/main.cpp(212) OutputPAF = 1
INFO: src/main.cpp(215) BinSize = 500
INFO: src/main.cpp(218) DeltaChernoff = 0.100000
INFO: src/main.cpp(221) RunPairwiseAlignment = 1
INFO: src/main.cpp(231) HOPC = DISABLED
INFO: src/main.cpp(235) xDrop = 7
INFO: src/main.cpp(238) KmerSplitCount = 1
INFO: src/main.cpp(243) useMinimizer = ENABLED
INFO: src/main.cpp(245) minimizerWindow = 17
INFO: src/main.cpp(276) numThreads = 64
INFO: src/../include/kmercount.hpp(711) ReadingFASTQ = metagenome-anonymous.fastq
*** Error in `bella': free(): invalid pointer: 0x00002aaacc076480 ***
Aborted

When I run without the "-w 17" option, it runs perfectly.

Thanks,
Gabe

Issue I neglected earlier

Data set:

/project/projectdirs/mp309/bella-spgemm/ecoli_hifi_29x.fastq

Parameters:

-k 31 -l 20 -u 23

Fix:

The issue appears to be in include/common/transpose.h, ~ line 35.

for (IT i=0; i <= n; i++)
{
    cscColPtr[i+1] = atomicColPtr[i] + cscColPtr[i];
}

cscColPtr

is of length n+1 I believe (n rows, n+1 stores nnz).

However this writes at location n+1 (i.e. n+2nd element), causing out of bounds.

I believe the correct code would be

for (IT i=0; i < n; i++)

cscColPtr[i+1] already handles the +1 size correctly for the last index (n).

Hi,
when building bella-gpu (make bella-gpu) ; I've got an error

nvcc -arch=sm_70 -O3 -maxrregcount=32 -std=c++14 -Xcompiler -fopenmp -w -Iinclude/common/GTgraph/sprng2.0-lite/include -IloganGPU -Iseqan -o bella hash_funcs.o Kmer.o Buffer.o fq_reader.o optlist.o src/main.cu -L/home/boelle/Documents/bella/libbloom/build -lbloom -lpthread -lbz2 -lz -D__NVCC__

seqan/seqan/score/score_matrix_dyn.h:110:489: error: template argument 1 is invalid
110 | enum class AminoAcidScoreMatrixID : std::underlying_type_t<decltype(Find<impl::score::MatrixTags, ScoreSpecBlosum30>::VALUE)>

Seems like some template library must be missing - Do you have an idea ?
building without GPU is ok (make bella ); building LOGAN runs fine

platform : Ubuntu 20.04
gcc : 9.4.0
nvcc : 11.6 (quadro P1000 - architecture=sm_62)

Thanks a lot.

Hi,

I was working with the evaluation module of the Bella. I was going to test the output from MECAT. So, there I need to provide two output files: one is Mecat's output file and the other one id Mecat's indices files. I am not sure what does that " Meact's indices" file mean? Where where to find it?