iqbal-lab-org / pandora Goto Github PK

View Code? Open in Web Editor NEW

109.0 109.0 14.0 287.08 MB

Pan-genome inference and genotyping with long noisy or short accurate reads

License: MIT License

CMake 1.86% C++ 95.51% Shell 0.91% Python 0.10% Dockerfile 0.14% C 1.48%

pandora's People

Contributors

Stargazers

Watchers

Forkers

pythseq ffranr mbhall88 bricoletc x-bioinformatics winterlich leoisl slw287r vikash84 kdm9 wook2014 schaudge

pandora's Issues

Clean up GATB graph h5 files after finished

GATB dumps a .h5 file for each graph it builds. These are not needed later and can clog up the file system so we should remove them at the end of each local assembly.

Boost linker error

Boost linker error when compiling ca57d6c.

/usr/local/bin/cmake --build /home/rffrancon/Documents/pandora/cmake-build-debug --target pandora -- -j 2
[ 20%] Built target boost
Scanning dependencies of target pandora
[ 25%] Building CXX object CMakeFiles/pandora.dir/src/check_kmergraph_main.cpp.o
[ 25%] Building CXX object CMakeFiles/pandora.dir/src/compare_main.cpp.o
[ 25%] Building CXX object CMakeFiles/pandora.dir/src/estimate_parameters.cpp.o
[ 28%] Building CXX object CMakeFiles/pandora.dir/src/fastaq_handler.cpp.o
[ 30%] Building CXX object CMakeFiles/pandora.dir/src/get_vcf_ref_main.cpp.o
[ 33%] Building CXX object CMakeFiles/pandora.dir/src/index.cpp.o
[ 35%] Building CXX object CMakeFiles/pandora.dir/src/index_main.cpp.o
/home/rffrancon/Documents/pandora/src/index_main.cpp: In function ‘int pandora_index(int, char**)’:
/home/rffrancon/Documents/pandora/src/index_main.cpp:54:10: warning: variable ‘update’ set but not used [-Wunused-but-set-variable]
     bool update = false;
          ^~~~~~
[ 38%] Building CXX object CMakeFiles/pandora.dir/src/interval.cpp.o
[ 41%] Building CXX object CMakeFiles/pandora.dir/src/inthash.cpp.o
[ 41%] Building CXX object CMakeFiles/pandora.dir/src/kmergraph.cpp.o
[ 43%] Building CXX object CMakeFiles/pandora.dir/src/kmernode.cpp.o
[ 46%] Building CXX object CMakeFiles/pandora.dir/src/localPRG.cpp.o
[ 48%] Building CXX object CMakeFiles/pandora.dir/src/localgraph.cpp.o
[ 51%] Building CXX object CMakeFiles/pandora.dir/src/localnode.cpp.o
[ 53%] Building CXX object CMakeFiles/pandora.dir/src/main.cpp.o
[ 56%] Building CXX object CMakeFiles/pandora.dir/src/map_main.cpp.o
[ 56%] Building CXX object CMakeFiles/pandora.dir/src/minihit.cpp.o
[ 58%] Building CXX object CMakeFiles/pandora.dir/src/minihits.cpp.o
[ 61%] Building CXX object CMakeFiles/pandora.dir/src/minimizer.cpp.o
[ 64%] Building CXX object CMakeFiles/pandora.dir/src/minirecord.cpp.o
[ 66%] Building CXX object CMakeFiles/pandora.dir/src/noise_filtering.cpp.o
[ 69%] Building CXX object CMakeFiles/pandora.dir/src/path.cpp.o
[ 71%] Building CXX object CMakeFiles/pandora.dir/src/seq.cpp.o
[ 71%] Building CXX object CMakeFiles/pandora.dir/src/utils.cpp.o
/home/rffrancon/Documents/pandora/src/utils.cpp: In function ‘void infer_localPRG_order_for_reads(const std::vector<LocalPRG*>&, MinimizerHits*, pangenome::Graph*, int, const uint&, const float&, uint, uint)’:
/home/rffrancon/Documents/pandora/src/utils.cpp:415:69: warning: unused parameter ‘genome_size’ [-Wunused-parameter]
                                     const int max_diff, const uint &genome_size, const float &scale_cluster_size,
                                                                     ^~~~~~~~~~~
[ 74%] Building CXX object CMakeFiles/pandora.dir/src/vcf.cpp.o
[ 76%] Building CXX object CMakeFiles/pandora.dir/src/vcfrecord.cpp.o
[ 79%] Building CXX object CMakeFiles/pandora.dir/src/walk_main.cpp.o
[ 82%] Building CXX object CMakeFiles/pandora.dir/src/de_bruijn/graph.cpp.o
[ 84%] Building CXX object CMakeFiles/pandora.dir/src/de_bruijn/node.cpp.o
[ 87%] Building CXX object CMakeFiles/pandora.dir/src/de_bruijn/ns.cpp.o
[ 87%] Building CXX object CMakeFiles/pandora.dir/src/pangenome/ns.cpp.o
[ 89%] Building CXX object CMakeFiles/pandora.dir/src/pangenome/pangraph.cpp.o
[ 92%] Building CXX object CMakeFiles/pandora.dir/src/pangenome/pannode.cpp.o
/home/rffrancon/Documents/pandora/src/pangenome/pannode.cpp:189:15: warning: ‘std::ostream& pangenome::operator<<(std::ostream&, const pangenome::Node&)’ has not been declared within pangenome
 std::ostream &pangenome::operator<<(std::ostream &out, pangenome::Node const &n) {
               ^~~~~~~~~
In file included from /home/rffrancon/Documents/pandora/src/pangenome/pannode.cpp:5:0:
/home/rffrancon/Documents/pandora/include/pangenome/pannode.h:43:26: note: only here as a friend
     friend std::ostream &operator<<(std::ostream &out, const Node &n);
                          ^~~~~~~~
/home/rffrancon/Documents/pandora/src/pangenome/pangraph.cpp:481:15: warning: ‘std::ostream& pangenome::operator<<(std::ostream&, const pangenome::Graph&)’ has not been declared within pangenome
 std::ostream &pangenome::operator<<(std::ostream &out, pangenome::Graph const &m) {
               ^~~~~~~~~
In file included from /home/rffrancon/Documents/pandora/include/utils.h:12:0,
                 from /home/rffrancon/Documents/pandora/src/pangenome/pangraph.cpp:12:
/home/rffrancon/Documents/pandora/include/pangenome/pangraph.h:77:26: note: only here as a friend
     friend std::ostream &operator<<(std::ostream &out, const Graph &m);
                          ^~~~~~~~
[ 94%] Building CXX object CMakeFiles/pandora.dir/src/pangenome/panread.cpp.o
/home/rffrancon/Documents/pandora/src/pangenome/panread.cpp:216:15: warning: ‘std::ostream& pangenome::operator<<(std::ostream&, const pangenome::Read&)’ has not been declared within pangenome
 std::ostream &pangenome::operator<<(std::ostream &out, const pangenome::Read &r) {
               ^~~~~~~~~
In file included from /home/rffrancon/Documents/pandora/src/pangenome/panread.cpp:10:0:
/home/rffrancon/Documents/pandora/include/pangenome/panread.h:41:26: note: only here as a friend
     friend std::ostream &operator<<(std::ostream &out, const Read &r);
                          ^~~~~~~~
[ 97%] Building CXX object CMakeFiles/pandora.dir/src/pangenome/pansample.cpp.o
/home/rffrancon/Documents/pandora/src/pangenome/pansample.cpp:37:15: warning: ‘std::ostream& pangenome::operator<<(std::ostream&, const pangenome::Sample&)’ has not been declared within pangenome
 std::ostream &pangenome::operator<<(std::ostream &out, pangenome::Sample const &s) {
               ^~~~~~~~~
In file included from /home/rffrancon/Documents/pandora/src/pangenome/pansample.cpp:7:0:
/home/rffrancon/Documents/pandora/include/pangenome/pansample.h:31:26: note: only here as a friend
     friend std::ostream &operator<<(std::ostream &out, const Sample &s);
                          ^~~~~~~~
[100%] Linking CXX executable pandora
/usr/bin/ld: /home/rffrancon/Documents/pandora/cmake-build-debug/lib/libboost_filesystem.a(operations.o): relocation R_X86_64_32 against `.rodata.str1.1' cannot be used when making a shared object; recompile with -fPIC
/usr/bin/ld: /home/rffrancon/Documents/pandora/cmake-build-debug/lib/libboost_filesystem.a(path.o): relocation R_X86_64_32 against `.rodata.str1.8' cannot be used when making a shared object; recompile with -fPIC
/usr/bin/ld: /home/rffrancon/Documents/pandora/cmake-build-debug/lib/libboost_system.a(error_code.o): relocation R_X86_64_32 against `.rodata.str1.1' cannot be used when making a shared object; recompile with -fPIC
/usr/bin/ld: final link failed: Nonrepresentable section on output
collect2: error: ld returned 1 exit status
CMakeFiles/pandora.dir/build.make:979: recipe for target 'pandora' failed
make[3]: *** [pandora] Error 1
CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/pandora.dir/all' failed
make[2]: *** [CMakeFiles/pandora.dir/all] Error 2
CMakeFiles/Makefile2:79: recipe for target 'CMakeFiles/pandora.dir/rule' failed
make[1]: *** [CMakeFiles/pandora.dir/rule] Error 2
Makefile:129: recipe for target 'pandora' failed
make: *** [pandora] Error 2

Estimate global covg better

To convert to an integer, round rather than implicit casting on line
https://github.com/rmcolq/pandora/blob/3df77ce343f2e1483f60a7188d98197209e92b18/src/utils.cpp#L461
Also add an assert which fails at this point if coverage is 0 - too low.

Implement invertible kmer hash function

Want to avoid poly-A kmers always achieving the minimizing value - distribute minimizing kmers more evenly across the set of all kmers

Here is a quick excel graph showing for some values of coverage (x-axis) the time, max memory (RAM in MB) and average memory used by pandora.

To be updated with a better graph and extended to 900x (since we have the dataset for it.
Note that max RAM seems to grow linearly with coverage.

Duplicate coordinates

In the function get_read_overlap_coordinates https://github.com/rmcolq/pandora/blob/a4a596cb17c319048f294b286301e2bcbe9ee17a/src/extract_reads.cpp#L157-L200 the output vector often has multiple instances of the exact same coordinates.
As we later loop over this vector of coordinates and do operations on reads based on them (https://github.com/rmcolq/pandora/blob/a4a596cb17c319048f294b286301e2bcbe9ee17a/src/extract_reads.cpp#L239-L301) maybe we should remove duplicates or use a set of some sort?

Remove temp directory from root

@rmcolq Can this folder be removed?
https://github.com/rffrancon/pandora/tree/dev/temp

Incorrect detection of non-ACGT base

When running pandora map on /nfs/leia/research/iqbal/mbhall/Projects/Pandora_variation/Analysis/Pandora/toy.fastq the log file shows that non-ACGT bases were found in multiple reads. On inspection of the reads in question, I am unable to find any non-ACGT bases.

Log file and fastq file attached below.

pandora.map.log
toy.fastq.gz

Acceptable local assembly performance

@mbhall88 has been running "latest" Pandora with the perf changes you 3 produced on the full graph with a cardio sample. So far running for days, sitting in the "get intervals" stage.
@mbhall88 feel free to edit/update with details.

Memory leaks when running unit tests

Valgrind output: valgrind_output.txt

Valgrind command used:
valgrind --log-file="valgrind_output" ../../cmake-build-debug/test/pandora_test
with working directory: pandora/test/test_cases. This command was executed on my laptop. All of the tests passed.

Commit: b914159
Branch: dev

For example:

==15934== Process terminating with default action of signal 6 (SIGABRT)
==15934==    at 0x5DF9E97: raise (raise.c:51)
==15934==    by 0x5DFB800: abort (abort.c:79)
==15934==    by 0x5DEB399: __assert_fail_base (assert.c:92)
==15934==    by 0x5DEB411: __assert_fail (assert.c:101)
==15934==    by 0x5E29E9: Interval::Interval(unsigned int, unsigned int) (interval.cpp:11)
==15934==    by 0x6136FD: Minimizer::Minimizer(unsigned long, unsigned int, unsigned int, bool) (minimizer.cpp:11)
==15934==    by 0x71C536: MinimizerTest_create_Test::TestBody() (minimizer_test.cpp:51)
==15934==    by 0x800F76: void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) (in /home/rffrancon/github/pandora/cmake-build-debug/test/pandora_test)
==15934==    by 0x7FAEA2: void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) (in /home/rffrancon/github/pandora/cmake-build-debug/test/pandora_test)
==15934==    by 0x7D99E9: testing::Test::Run() (in /home/rffrancon/github/pandora/cmake-build-debug/test/pandora_test)
==15934==    by 0x7DA346: testing::TestInfo::Run() (in /home/rffrancon/github/pandora/cmake-build-debug/test/pandora_test)
==15934==    by 0x7DA9C9: testing::TestCase::Run() (in /home/rffrancon/github/pandora/cmake-build-debug/test/pandora_test)
==15934== 
==15934== HEAP SUMMARY:
==15934==     in use at exit: 7,154,922 bytes in 24,129 blocks
==15934==   total heap usage: 571,866 allocs, 547,737 frees, 125,362,426,822 bytes allocated
==15934== 
==15934== LEAK SUMMARY:
==15934==    definitely lost: 5,792 bytes in 71 blocks
==15934==    indirectly lost: 2,722,572 bytes in 14,408 blocks
==15934==      possibly lost: 1,996,141 bytes in 931 blocks
==15934==    still reachable: 2,430,417 bytes in 8,719 blocks
==15934==         suppressed: 0 bytes in 0 blocks
==15934== Rerun with --leak-check=full to see details of leaked memory

Fix `hits_along_path`

For a test case with lmp (1 [4, 5) G) (2 [8, 9) C)
It doesn't think hit (3, 3, 3, 3{[4, 5)[8, 9)[16, 17)}, 1, 0) lies on this path

Fix `kmernode_path_from_localnode_path`

Found a wrong answer in function:

    uint32_t pnode_id = 3, prg_id = 3, read_id = 0, knode_id = 0;
    string pnode_name = "three";
    bool orientation(true);
    deque<Interval> d;
    prg::Path prg_path;
    MinimizerHitPtr mh;

    // define localPRG
    LocalPRG l3(prg_id, "nested varsite", "A 5 G 7 C 8 T 7 T 9 CCG 10 CGG 9  6 G 5 TAT");
    Index *idx;
    idx = new Index();
    auto w = 1, k = 3;
    l3.minimizer_sketch(idx, w, k);

    // define localpath and kmerpath
    // corresponds to sequence (A) G C T CGG  (TAT)
    vector<LocalNodePtr> lmp = {//l3.prg.nodes[0],
            l3.prg.nodes[1], l3.prg.nodes[2], l3.prg.nodes[4],
            l3.prg.nodes[6], l3.prg.nodes[7]//, l3.prg.nodes[9]
    };
    for (const auto &l : lmp)
        cout << *l << endl;

    vector<KmerNodePtr> kmp = l3.kmernode_path_from_localnode_path(lmp);
    for (const auto &n : kmp)
        cout << *n << endl;

Finds an additional kmernode 3 3{[0, 1)[36, 37)[40, 41)} 0, 0 which it shouldn't

Change path termination process for local assembly

Currently, as we build a path I test whether it ends in the end k-mer. But it would make more sense to test whether it ends in any of the end k-mers we have which exist in the graph.

To test which end k-mers exist we can use the hash/method we produce for #59 to mark all end k-mers that exist in the graph upfront (and their reverse).

Clean up CLI parameter parsing

Robyn suggested this as a cleaner way to have parameters and help messages

Default maximum path length for local assembly not being overwritten

When setting the max_path_length for local assembly, such as here, the value seems to always come through in the log file as 30 (which is the default set here).

Optimise `get_node` function

get_node is currently causing up to 6 iterations through the graph per start/end kmer combination.

Optimize index construction

Speed up update_covgs

Currently takes about 10 mins for a single nanopore read, and 1 PRG, everything else takes seconds.

Decide if LocalPRG::find_path_and_variants is redundant and can be removed

https://github.com/rmcolq/pandora/blob/3ee4d077ada2e99d5f322e29d874be530c7bab67/src/localPRG.cpp#L1414

Create separate function to run local_assembly on an in memory fasta including all gene read pileups

Add a function to sort a fastaq by the header strings

Remove file writes during tests

Change code so use sstream and can compare these during test runs without writing to file

Implement boost logging

Add test for set_p happening in estimate_parameters.cpp

Restructure local assembly function

Remove goto statements and replace with a function call.

Apply save_read_strings_to_denovo_assemble to a pangraph rather than on a per-gene basis

Rewrite body of extract_reads.cpp to reduce complexity and looping

Make sure compare command runs

Add tests (multisample dataset)
Update in line with how map now works

include genotyping
updates which fix the localPRG as something that doesn't change and creates a copy of kmergraphs within pangraph structure where counts/coverages are added
see how new vcf structure works with adding samples/reads in this case

Add test for remove_shortcut_edges() in kmergraph.cpp

Update README

probably good to do

Fix anything with vcf_ref

Function get_vcf_ref sometimes outputs vcfs which are far too short to be real. Also must output some other sort of bad vcf (not yet diagnosed).
Latest fail on gene GC00000258_r1_1 in vcf_ref /nfs/leia/research/iqbal/rmcolq/projects/pangenome_prg/ecoli/290818_all/ecoli_pangenome_PRG_290818.k12_vcf_ref.fa in directory /nfs/leia/research/iqbal/rmcolq/projects/pandora_variation/analysis/loman_k12_debug/work/5b/be999de3e77a9778ffc7f449005cee/.

In the very least, asserts need to be added to catch bad vcf_refs.

If no nodes found, output error message to reduce genome_size

If at the end of the loop through pangraph nodes there are no nodes due to this being called
https://github.com/rmcolq/pandora/blob/3df77ce343f2e1483f60a7188d98197209e92b18/src/map_main.cpp#L285
Then output a message querying if genome_size is correct. Default was 5,000,000

Create updated PRG file for the subset of genes found present (including genes found exactly, and those with denovo alleles to be included)

Please improve title.

What should the license be?

Should add a license to this repo

Compare command outputs single VCF instead of multiple directories

Currently outputs a pangenome matrix file with a dodgy name (missing /?)
Currently outputs a directory for every gene found in any sample - make this one for everything.

Handle rev complements

Note to self

uint errors when compiling with gcc > 4.2 on Mac

Re: comments from @rffrancon in #15 all occurrences of uint should be changed to uint32_t for consistency.

Clear up github branches

Memory leaks

Check the output from command:
/home/zam/bin/bin/valgrind --leak-check=yes --track-origins=yes -v build/pandora test/test_cases/prg4567.fa test/test_cases/read0.fa wow &> build/pandora_valgrind_output.txt

START: Wed Oct 26 20:23:44 2016

Building Index from PRG file
--39147-- REDIR: 0x3a146bd1d0 (operator new[](unsigned long)) redirected to 0x4a0794a (operator new[](unsigned long))
--39147-- REDIR: 0x3a11e83350 (memchr) redirected to 0x4a09890 (memchr)
--39147-- REDIR: 0x3a146bb2f0 (operator delete(void_)) redirected to 0x4a072ec (operator delete(void_))
--39147-- REDIR: 0x3a11e833d0 (bcmp) redirected to 0x480260a (_vgnU_ifunc_wrapper)
--39147-- REDIR: 0x3a11f3e6d0 (__memcmp_sse4_1) redirected to 0x4a0a8c0 (bcmp)
--39147-- REDIR: 0x3a11e83800 (memmove) redirected to 0x4a0ad40 (memmove)
--39147-- REDIR: 0x3a146bb330 (operator delete) redirected to 0x4a06f68 (operator delete)
Reading Read file
Finished building PanGraph from reads
Writing PanGraph to file
FINISH: Wed Oct 26 20:23:46 2016

==39147==
==39147== HEAP SUMMARY:
==39147== in use at exit: 504 bytes in 16 blocks
==39147== total heap usage: 1,934 allocs, 1,918 frees, 454,491 bytes allocated
==39147==
==39147== Searching for pointers to 16 not-freed blocks
==39147== Checked 207,608 bytes
==39147==
==39147== 504 (48 direct, 456 indirect) bytes in 1 blocks are definitely lost in loss record 9 of 9
==39147== at 0x4A07EB6: operator new(unsigned long) (vg_replace_malloc.c:287)
==39147== by 0x44790C: main (in /data2/users/rachel/projects/pandora/build/pandora)
==39147==
==39147== LEAK SUMMARY:
==39147== definitely lost: 48 bytes in 1 blocks
==39147== indirectly lost: 456 bytes in 15 blocks
==39147== possibly lost: 0 bytes in 0 blocks
==39147== still reachable: 0 bytes in 0 blocks
==39147== suppressed: 0 bytes in 0 blocks
==39147==
==39147== ERROR SUMMARY: 3 errors from 3 contexts (suppressed: 4 from 4)
==39147==
==39147== 1 errors in context 1 of 3:
==39147== Conditional jump or move depends on uninitialised value(s)
==39147== at 0x3A11A1754B: index (in /lib64/ld-2.12.so)
==39147== by 0x3A11A06237: expand_dynamic_string_token (in /lib64/ld-2.12.so)
==39147== by 0x3A11A082F1: _dl_map_object (in /lib64/ld-2.12.so)
==39147== by 0x3A11A016DD: map_doit (in /lib64/ld-2.12.so)
==39147== by 0x3A11A0E265: _dl_catch_error (in /lib64/ld-2.12.so)
==39147== by 0x3A11A015F6: do_preload (in /lib64/ld-2.12.so)
==39147== by 0x3A11A0480E: dl_main (in /lib64/ld-2.12.so)
==39147== by 0x3A11A15B4D: _dl_sysdep_start (in /lib64/ld-2.12.so)
==39147== by 0x3A11A014A3: _dl_start (in /lib64/ld-2.12.so)
==39147== by 0x3A11A00B07: ??? (in /lib64/ld-2.12.so)
==39147== by 0x3: ???
==39147== by 0xFFEFFF912: ???
==39147== Uninitialised value was created by a stack allocation
==39147== at 0x3A11A032A1: dl_main (in /lib64/ld-2.12.so)
==39147==
==39147==
==39147== 1 errors in context 2 of 3:
==39147== Conditional jump or move depends on uninitialised value(s)
==39147== at 0x3A11A17546: index (in /lib64/ld-2.12.so)
==39147== by 0x3A11A06237: expand_dynamic_string_token (in /lib64/ld-2.12.so)
==39147== by 0x3A11A082F1: _dl_map_object (in /lib64/ld-2.12.so)
==39147== by 0x3A11A016DD: map_doit (in /lib64/ld-2.12.so)
==39147== by 0x3A11A0E265: _dl_catch_error (in /lib64/ld-2.12.so)
==39147== by 0x3A11A015F6: do_preload (in /lib64/ld-2.12.so)
==39147== by 0x3A11A0480E: dl_main (in /lib64/ld-2.12.so)
==39147== by 0x3A11A15B4D: _dl_sysdep_start (in /lib64/ld-2.12.so)
==39147== by 0x3A11A014A3: _dl_start (in /lib64/ld-2.12.so)
==39147== by 0x3A11A00B07: ??? (in /lib64/ld-2.12.so)
==39147== by 0x3: ???
==39147== by 0xFFEFFF912: ???
==39147== Uninitialised value was created by a stack allocation
==39147== at 0x3A11A032A1: dl_main (in /lib64/ld-2.12.so)
==39147==
--39147--
--39147-- used_suppression: 4 U1004-ARM-_dl_relocate_object
==39147==
==39147== ERROR SUMMARY: 3 errors from 3 contexts (suppressed: 4 from 4)

Store gene, interval and sub_lmp string in an object

fix debruijn graph so can have more than 32000 PRGs

Have been using a uint16_t which is too small

Catch if dodgy fastq input

Add assert that read sequence and quality lines are the same length
Is it correct that spec is
@Header
SEQ
+
QUA

Standardise the use of boost::filesystem for directories and filepaths

Optimise writes

Use \n rather than endl and write longer strings less often, based on Martin's advice about the performance hit.

Investigating GATB for de-Bruijn graph analysis

I'm not sure how to properly install GATB so that we can use it as a library. I've opened an issue here: GATB/gatb-core#21

This is the work that we've done so far on this issue: mbhall88@346c4f7

Save mapped read strings and coordinates takes forever

In src/pangenome/pangraph.cpp I now have a function:

void Graph::save_mapped_read_strings(const string& readfilepath, \\
const string& outdir, const int32_t buff)

Which for each gene, finds the coordinates on the reads for each cluster of hits and creates a file with this segment of read output to it.
This function takes about 2 days (or more) to run on full K12 pangenome and 900X nanopore reads (compared to an hour and a half for the rest of the map code)

One thing we can do to add value to the best-reference-per-gene that Pandora infers, is simply to annotate it. We could just run Prokka

https://github.com/tseemann/prokka

and since we have a file with one read per gene, I think it would be super easy.

segfault if index contains different number of prgs to prgfile

make a clear error message