easy way to dump kmers with certain prefixes?

[chunlinxiao]

Hi,

is there any fast way that kmc just dump kmers with certain prefixes, other than looking through the whole kmc database?

thanks

chunlin

[marekkokot]

Hi,

It would be possible, but currently, our software does not support such an option. Nevertheless, you may implement it on your own (unfortunately it is rather not trivial). K-mers in kmc database (*.kmc_pre, *.kmc_suf files) are distributed among a number of (512 in default mode) bins. Each bin contains k-mers in sorted order (they are stores using additional LUT table to save a space and speed up search), so a binary search may be used to find boundaries of k-mers with specified prefixes in each bin. In some cases, however, k-mers are not distributed to bins (if kmc database is an output of most of kmc_tools operations, or when kmc was run with small k (about 14 or less)). There is a value that informs which case is in the given database.

You may always convert the most complex case of k-mers distributed to bins to simpler when all k-mers are sorted using kmc_tools sort operation.

The binary format of the database is described in this document: https://github.com/refresh-bio/KMC/blob/master/API.pdf In the case when all k-mers are sorted and not distributed to bins the database format is described here: https://static-content.springer.com/esm/art%3A10.1186%2F1471-2105-14-160/MediaObjects/12859_2012_5911_MOESM1_ESM.pdf

If you decide to implement it, you may use kmc_dump source code as a reference, because it supports both database formats. In general, kmc_dump uses kmc_api, whose source codes is avaiable here: https://github.com/refresh-bio/KMC/tree/master/kmc_api

You may also use kmc_tools source code as a reference, but it is more coplex and will require more time, but kmc_tools dump operation is a little faster than kmc_dump.

In case of any questions do not hesitate and ask.

We will consider adding support for dump operation witthe h specified prefix, but unfortunatelly even if we decide to implement it, it will be for sure not in the near future.

[chunlinxiao]

thank you marekkokot very much for the detailed reply.

I just installed your kmc3.1.1 successfully from the source codes. But when I tried to compile kmc_dump_sample.cpp under directory of kmc_dump_sample using g++ ( 5.4.0 ), I have the following error:

g++ kmc_dump_sample.cpp

In file included from kmc_dump_sample.cpp:16:0: ../kmc_api/kmc_file.h:90:8: error: expected nested-name-specifier before ‘super_kmers_t’ using super_kmers_t = std::vector<std::tuple<uint32, uint32, uint32>>;//start_pos, len, bin_n ^ ../kmc_api/kmc_file.h:91:58: error: ‘super_kmers_t’ has not been declared void GetSuperKmers(const std::string& transformed_read, super_kmers_t& super_kmers);

did I miss anything?

thanks

[marekkokot]

Hi, probably you should also specify -std=c++11, but there may be some missing references. Try to replace makefile content in the main directory with:

all: kmc
	
KMC_BIN_DIR = bin
KMC_MAIN_DIR = kmer_counter
KMC_API_DIR = kmc_api
KMC_DUMP_DIR = kmc_dump
KMC_DUMP_SAMPLE_DIR = kmc_dump_sample
KMC_TOOLS_DIR = kmc_tools

CC 	= g++
CFLAGS	= -Wall -O3 -m64 -static -Wl,--whole-archive -lpthread -Wl,--no-whole-archive -std=c++11 
CLINK	= -lm -static -O3 -Wl,--whole-archive -lpthread -Wl,--no-whole-archive -std=c++11 

KMC_TOOLS_CFLAGS	= -Wall -O3 -m64 -static -Wl,--whole-archive -lpthread -Wl,--no-whole-archive -std=c++14
KMC_TOOLS_CLINK	= -lm -static -O3 -Wl,--whole-archive -lpthread -Wl,--no-whole-archive -std=c++14

DISABLE_ASMLIB = false

KMC_OBJS = \
$(KMC_MAIN_DIR)/kmer_counter.o \
$(KMC_MAIN_DIR)/mmer.o \
$(KMC_MAIN_DIR)/mem_disk_file.o \
$(KMC_MAIN_DIR)/rev_byte.o \
$(KMC_MAIN_DIR)/bkb_writer.o \
$(KMC_MAIN_DIR)/cpu_info.o \
$(KMC_MAIN_DIR)/bkb_reader.o \
$(KMC_MAIN_DIR)/fastq_reader.o \
$(KMC_MAIN_DIR)/timer.o \
$(KMC_MAIN_DIR)/develop.o \
$(KMC_MAIN_DIR)/kb_completer.o \
$(KMC_MAIN_DIR)/kb_storer.o \
$(KMC_MAIN_DIR)/kmer.o \
$(KMC_MAIN_DIR)/prob_qual.o
RADULS_OBJS = \
$(KMC_MAIN_DIR)/raduls_sse2.o \
$(KMC_MAIN_DIR)/raduls_sse41.o \
$(KMC_MAIN_DIR)/raduls_avx2.o \
$(KMC_MAIN_DIR)/raduls_avx.o 

KMC_LIBS = \
$(KMC_MAIN_DIR)/libs/libz.a \
$(KMC_MAIN_DIR)/libs/libbz2.a

KMC_DUMP_OBJS = \
$(KMC_DUMP_DIR)/nc_utils.o \
$(KMC_DUMP_DIR)/kmc_dump.o 

KMC_DUMP_SAMPLE_OBJS = \
$(KMC_DUMP_SAMPLE_DIR)/kmc_dump_sample.o

KMC_API_OBJS = \
$(KMC_API_DIR)/mmer.o \
$(KMC_API_DIR)/kmc_file.o \
$(KMC_API_DIR)/kmer_api.o

KMC_TOOLS_OBJS = \
$(KMC_TOOLS_DIR)/kmc_header.o \
$(KMC_TOOLS_DIR)/kmc_tools.o \
$(KMC_TOOLS_DIR)/nc_utils.o \
$(KMC_TOOLS_DIR)/parameters_parser.o \
$(KMC_TOOLS_DIR)/parser.o \
$(KMC_TOOLS_DIR)/tokenizer.o \
$(KMC_TOOLS_DIR)/fastq_filter.o \
$(KMC_TOOLS_DIR)/fastq_reader.o \
$(KMC_TOOLS_DIR)/fastq_writer.o \
$(KMC_TOOLS_DIR)/percent_progress.o

KMC_TOOLS_LIBS = \
$(KMC_TOOLS_DIR)/libs/libz.a \
$(KMC_TOOLS_DIR)/libs/libbz2.a 

ifeq ($(DISABLE_ASMLIB),true)
	CFLAGS += -DDISABLE_ASMLIB
	KMC_TOOLS_CFLAGS += -DDISABLE_ASMLIB
else
	KMC_LIBS += \
	$(KMC_MAIN_DIR)/libs/libaelf64.a 
	KMC_TOOLS_LIBS += \
	$(KMC_TOOLS_DIR)/libs/libaelf64.a 
endif 	

$(KMC_OBJS) $(KMC_DUMP_OBJS) $(KMC_API_OBJS) $(KMC_DUMP_SAMPLE_OBJS): %.o: %.cpp
	$(CC) $(CFLAGS) -c $< -o $@

$(KMC_TOOLS_OBJS): %.o: %.cpp
	$(CC) $(KMC_TOOLS_CFLAGS) -c $< -o $@

$(KMC_MAIN_DIR)/raduls_sse2.o: $(KMC_MAIN_DIR)/raduls_sse2.cpp
	$(CC) $(CFLAGS) -msse2 -c $< -o $@
$(KMC_MAIN_DIR)/raduls_sse41.o: $(KMC_MAIN_DIR)/raduls_sse41.cpp
	$(CC) $(CFLAGS) -msse4.1 -c $< -o $@
$(KMC_MAIN_DIR)/raduls_avx.o: $(KMC_MAIN_DIR)/raduls_avx.cpp
	$(CC) $(CFLAGS) -mavx -fabi-version=0 -c $< -o $@
$(KMC_MAIN_DIR)/raduls_avx2.o: $(KMC_MAIN_DIR)/raduls_avx2.cpp
	$(CC) $(CFLAGS) -mavx2 -mfma -fabi-version=0 -c $< -o $@
$(KMC_MAIN_DIR)/instrset_detect.o: $(KMC_MAIN_DIR)/libs/vectorclass/instrset_detect.cpp
	$(CC) $(CFLAGS) -c $< -o $@
	
kmc: $(KMC_OBJS) $(RADULS_OBJS) $(KMC_MAIN_DIR)/instrset_detect.o 
	-mkdir -p $(KMC_BIN_DIR)
	$(CC) $(CLINK) -o $(KMC_BIN_DIR)/$@ $^ $(KMC_LIBS)
	
kmc_dump: $(KMC_DUMP_OBJS) $(KMC_API_OBJS)
	-mkdir -p $(KMC_BIN_DIR)
	$(CC) $(CLINK) -o $(KMC_BIN_DIR)/$@ $^

kmc_dump_sample: $(KMC_DUMP_SAMPLE_OBJS) $(KMC_API_OBJS)
	-mkdir -p $(KMC_BIN_DIR)
	$(CC) $(CLINK) -o $(KMC_BIN_DIR)/$@ $^
	
kmc_tools: $(KMC_TOOLS_OBJS) $(KMC_API_OBJS)
	-mkdir -p $(KMC_BIN_DIR)
	$(CC) $(KMC_TOOLS_CLINK) -o $(KMC_BIN_DIR)/$@ $^ $(KMC_TOOLS_LIBS)
	
clean:
	-rm $(KMC_MAIN_DIR)/*.o
	-rm $(KMC_API_DIR)/*.o
	-rm $(KMC_DUMP_DIR)/*.o
	-rm $(KMC_TOOLS_DIR)/*.o
	-rm $(KMC_DUMP_SAMPLE_DIR)/*o
	-rm -rf bin
	


all: kmc kmc_dump kmc_tools kmc_dump_sample

and than run:

make kmc_dump_sample

Let me know if it helps.

[chunlinxiao]

thank you very much - your new makefile works !

Before this, I did try $ g++ kmc_dump_sample.cpp -std=c++11

but it would give me the error with "error: ld returned 1 exit status".

I also tried the following with additional options (from your makefile) - similar errors below - do you have any suggestion for command line compilation ?

$ g++ kmc_dump_sample.cpp -lm -static -O3 -Wl,--whole-archive -lpthread -Wl,--no-whole-archive -std=c++11 -o test

kmc_dump_sample.cpp:(.text.startup+0x69): undefined reference to CKMCFile::CKMCFile()' kmc_dump_sample.cpp:(.text.startup+0x18c): undefined reference to CKMCFile::~CKMCFile()' kmc_dump_sample.cpp:(.text.startup+0x265): undefined reference to CKMCFile::OpenForListing(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)' kmc_dump_sample.cpp:(.text.startup+0x2b5): undefined reference to CKMCFile::Info(unsigned int&, unsigned int&, unsigned int&, unsigned int&, unsigned int&, unsigned int&, unsigned long long&, unsigned long long&)' kmc_dump_sample.cpp:(.text.startup+0x39e): undefined reference to CKMCFile::ReadNextKmer(CKmerAPI&, float&)' kmc_dump_sample.cpp:(.text.startup+0x436): undefined reference to CKmerAPI::char_codes' kmc_dump_sample.cpp:(.text.startup+0x479): undefined reference to CKmerAPI::char_codes' kmc_dump_sample.cpp:(.text.startup+0x49c): undefined reference to CKmerAPI::char_codes' kmc_dump_sample.cpp:(.text.startup+0x4c0): undefined reference to CKmerAPI::char_codes' kmc_dump_sample.cpp:(.text.startup+0x539): undefined reference to CKmerAPI::char_codes' /tmp/ccwLOWWX.o:kmc_dump_sample.cpp:(.text.startup+0x55c): more undefined references to CKmerAPI::char_codes' follow /tmp/ccwLOWWX.o: In function main': kmc_dump_sample.cpp:(.text.startup+0x5df): undefined reference to CKMCFile::ReadNextKmer(CKmerAPI&, unsigned int&)' kmc_dump_sample.cpp:(.text.startup+0x66e): undefined reference to CKmerAPI::char_codes' kmc_dump_sample.cpp:(.text.startup+0x6f8): undefined reference to CKMCFile::Close()' kmc_dump_sample.cpp:(.text.startup+0x740): undefined reference to CKMCFile::SetMaxCount(unsigned int)' kmc_dump_sample.cpp:(.text.startup+0x774): undefined reference to CKMCFile::SetMinCount(unsigned int)' kmc_dump_sample.cpp:(.text.startup+0x7c5): undefined reference to CKMCFile::~CKMCFile()' collect2: error: ld returned 1 exit status

[marekkokot]

This is because in your command line you do not compile all necessary cpp files. If you really need to compile this way (which I do not recommend) use:

g++ -O3 -std=c++14 kmc_dump_sample.cpp ../kmc_api/*.cpp

Let me know if it works.

[chunlinxiao]

yes it works fine - thank you very much Marek !