Running rMATS on HPC

[FarzanehRah]

Hi,

I ran rMATS on an HPC cluster, and there are two versions available as modules: rmats/4.1.2 and rmats/4.1.1. I also installed v4.3.0 from GitHub on the cluster (our clusters do not support conda). However, I’m encountering the same issue with all three versions:

traceback (most recent call last):
  File "~/rmats-turbo-master/rMATS_P/FDR.py", line 53, in <module>
    ifile = open(sys.argv[1]); title = ifile.readline()
FileNotFoundError: [Errno 2] No such file or directory: '../../output_11_24_comparison2/tmp/JC_SE/rMATS_result_P-V.txt'
paste: ../../output_11_24_comparison2/tmp/JC_SE/rMATS_result_FDR.txt: No such file or directory
........

I tried using the same versions of dependencies mentioned in the documentation, but for some, I wasn't able to use the specified versions (e.g., python/3.8.10 instead of python/3.6.12, proj4-fortran/1.0 ( proj4-fortran is a f77 and f90 wrappers for proj4) instead of gfortran (Fortran 77)).
I am getting this error when running the following command (v4.3.0):

gdb ~/rmats-turbo-master/rMATS_C/rMATSexe 
r -i ../../output_11_24_comparison2/JC.raw.input.SE.txt -t 1 -o ../../output_11_24_comparison2/tmp/JC_SE/rMATS_result_P-V.txt -c 0.0001

~/rmats-turbo-master/rMATS_C/rMATSexe: error while loading shared libraries: libgfortran.so.3: cannot open shared object file: No such file or directory
[Inferior 1 (process 1718094) exited with code 0177]

Is there any solution to this on HPC?

Thank you!

[EricKutschera]

From ~/rmats-turbo-master/rMATS_C/rMATSexe: error while loading shared libraries: libgfortran.so.3: cannot open shared object file: No such file or directory
It looks like you were able to compile rMATSexe and it is looking for a specific version of libgfortran.so. The build uses -lgfortran and I think it would have needed to find libgfortran.so somewhere on your system for the build to succeed: https://github.com/Xinglab/rmats-turbo/blob/v4.3.0/rMATS_C/Makefile#L12
Maybe you have libgfortran.so.3 and just need to set LD_LIBRARY_PATH to let it find it

You could try installing gfortran and then building from the source code again: https://gcc.gnu.org/wiki/GFortran

[FarzanehRah]

Thank you so much for your quick reply, I really appreciate it. I will try installing gfortran and building rmats from source code again.

[FarzanehRah]

Hi Eric,
I was able to resolve the initial error related to libgfortran.so.3 (i.e., the library loading issue), I am still encountering a segmentation fault when I try to run rmats.

I’ve tried debugging the issue using gdb and valgrind, but I haven’t been able to resolve it. valgrind suggests that there might be memory-related issues, but I cannot pinpoint the exact cause. Below is the output, there might be an issue related to memory handling or an incompatibility with the HPC environment?

gdb /cvmfs/soft.computecanada.ca/easybuild/software/2020/avx2/Compiler/gcc9/rmats/4.1.2/rMATS_C/rMATSexe

run ~/easybuild/software/2020/avx2/Compiler/gcc9/rmats/4.1.2/rMATS_C/rMATSexe -i ../a_test/JC.raw.input.SE.txt -o ../a_test/tmp/JC_SE/rMATS_result_P-V.txt -c 0.0001
[Thread debugging using libthread_db enabled]
Using host libthread_db library "~/gentoo/2020/lib64/libthread_db.so.1".
number of thread=1; input file=../a_test/JC.raw.input.SE.txt; output folder=../a_test/tmp/JC_SE/rMATS_result_P-V.txt; cutoff=0.0001;
Testing 189

Program received signal SIGSEGV, Segmentation fault.
0x000015554ba5f431 in mkl_blas_def_xdcopy () from ~/easybuild/software/2020/Core/imkl/2020.1.217/mkl/lib/intel64/libmkl_def.so`

valgrind --leak-check=full --track-origins=yes --show-reachable=yes --verbose ~/easybuild/software/2020/avx2/Compiler/gcc9/rmats/4.1.2/rMATS_C/rMATSexe -i ../a_test/JC.raw.input.SE.txt -o ../a_test/tmp/JC_SE/rMATS_result_P-V.txt -c 0.0001

valgrind: m_mallocfree.c:305 (get_bszB_as_is): Assertion 'bszB_lo == bszB_hi' failed.
valgrind: Heap block lo/hi size mismatch: lo = 112, hi = 3545286390305929524.
This is probably caused by your program erroneously writing past the
end of a heap block and corrupting heap metadata.  If you fix any
invalid writes reported by Memcheck, this assertion failure will
probably go away.  Please try that before reporting this as a bug.

host stacktrace:
--2476336-- VALGRIND INTERNAL ERROR: Valgrind received a signal 7 (SIGBUS) - exiting
--2476336-- si_code=128;  Faulting address: 0x0;  sp: 0x1008ca87a0

valgrind: the 'impossible' happened:
   Killed by fatal signal

host stacktrace:
Segmentation fault (core dumped)

Any advice or possible fixes would be greatly appreciated.
Thank you

[EricKutschera]

The error message from gdb shows an error with a blas library call, but it doesn't show where in the rmats code it happens

Program received signal SIGSEGV, Segmentation fault.
0x000015554ba5f431 in mkl_blas_def_xdcopy () from ~/easybuild/software/2020/Core/imkl/2020.1.217/mkl/lib/intel64/libmkl_def.so`

You could try compiling with debugging info: https://github.com/Xinglab/rmats-turbo/blob/v4.3.0/rMATS_C/Makefile#L13

change -O2 to -O0 -ggdb and recompile

Then if you run in gdb like before, when it shows the segfault you can run bt to get a back trace. Maybe you'll also be able to get gdb to show the variable values in the line of rmats code that made the library call

[FarzanehRah]

I was not able to figure out the issue on the HPC clusters, but successfully ran rMATS on Mac.
Thanks again for your time and help!