Delphi failure

[varnold2]

After running step3.py the following error occurs: WARNING: Delphi failed at focusing depth 1 of NTG01A0001_001, retry

WARNING: Delphi failed at focusing depth 1 of NTG01A0001_001, retry FATAL: too many failed delphi runs (10), quitting... Fatal error reported by conf_energies(), QUIT.

I checked run.prm to make sure it was going to correct directory and it is, the MCCE wiki stated that it could potentially be a rare bug and the crash would be harder to fix. Was wondering if the details for the fix are available.

EDIT: after searching through closed tickets, I attempted to run Delphi and it appears I am missing libgfortran.so.4, Are there any places available to get it?

[CatChenal]

Same here: I am using mcce installed locally in my conda environment (mce).

Using default_run.sh (the script used in benchmarking that bundles the first four mcce steps): mcce fails at step3 (for all proteins in a test set of size 4):

(mce) cat:~/projects/bench_tests/four_pdbs/clean_pdbs$ grep 'WARNING: Delphi' ./*/run.log | uniq
./135L/run.log:   WARNING: Delphi failed at focusing depth 1 of NTR01A0001_001, retry
./1A2P/run.log:   WARNING: Delphi failed at focusing depth 1 of NTR01A0003_001, retry
./1A6K/run.log:   WARNING: Delphi failed at focusing depth 1 of NTR01A0001_001, retry
./1A6M/run.log:   WARNING: Delphi failed at focusing depth 1 of NTR01A0001_001, retry

What I tried:

1. Fixing the delphi path in step3 did not change anything.
2. I then ran step3.py -r: head3.lst and step3_out.pdbs were created: that was a surprise!
3. I then launched step 4 and the pK.out file created looks fine.

As per @varnold2, I found out I actually cannot run delphi:

(mce) cat:~/projects/bench_tests/four_pdbs/clean_pdbs/1A6K$ delphi
delphi: error while loading shared libraries: libgfortran.so.4: cannot open shared object file: No such file or directory

PS: I did not have this issue on the server (isis): on isis I have my mcce path set in my .bashrc file to the "all users" version (/home/mcce...). Could it be that the server version and that in the newbooks conda channel diverge?

[varnold2]

Unsure if this will provide a solution for everyone but was able to run Delphi and step3.py by reverting to an older version of Ubuntu- specifically Ubuntu 18.04.6 LTS.

[CatChenal]

Thanks for your update, @varnold2.

I have found this, which I might try: https://ubuntu.pkgs.org/20.04/ubuntu-universe-arm64/libgfortran4_7.5.0-6ubuntu2_arm64.deb.html

[mbatgh]

for me installing mcce with conda (in my case actually micromamba), followed by installing libgfortran4 directly into the conda environment did the trick.

micromamba install libgfortran4

for some strange reason i had to set the PATH manually after that: echo PATH=/home/name/micromamba/pkgs/mcce-1.1.0-h3218e01_0/bin/:$PATH

[CatChenal]

Thanks for the feedback on your successful installation, @mbatgh. Could you please indicate your OS and its version. Thank you.

[mbatgh]

OS: Debian Trixie, Compilers, Fortran and C: gcc 13.2.0 with libgfortran in the conda environment replaced as shown above.

[CatChenal]

Thanks!

[Kivean]

I had the same issue on Ubuntu 22.04 LTS installing mcce through conda. Conda couldn't solve the mcce environment if I tried to manually install libgfortran4, apparently an incompatible set of dependencies. Looks like either newbooks or mcce is installing a different libgfortran version, not sure which.

Hilariously, I just rebuilt the mcce environment and installed libgfortran4 first, then mcce and newbooks, and that worked.