Guidance on Modifying AutoDock-GPU for Docking Large Molecules

[hezhizhang-269]

Hello,

I am interested in docking large molecules with AutoDock-GPU. As mentioned in #205, I modified the header file and it is working.

However, I noticed that there are other variables in the code with a default value of 256. I want to confirm whether modifying only MAX_NUM_OF_ATOMS is sufficient, or if I need to change these other variables as well to ensure proper functionality for larger molecules. While I understand that AutoDock-GPU is optimized for docking small molecules, are there any additional lines of code or parameters I should adjust to improve compatibility and performance for larger molecules?

Thank you for your guidance!

[diogomart]

Hello, perhaps the number of rotatable bonds too.

[hezhizhang-269]

Hi @diogomart,

I increased MAX_NUM_OF_ATOMS to 512 to dock the 1s40.zip pair, and it worked perfectly. However, when I further increased MAX_NUM_OF_ATOMS to 1024 (to accommodate other structure pairs that exceed 512 atoms), I encountered an issue. Testing with the same 1s40 pair now results in a segmentation fault, with no additional details provided.

Here’s the command I ran and the output I received:

caomq@cscigpu06:~/1s40$ autodock_gpu_128wi_1024_max --ffile 1s40_Protein_Model_1.maps.fld --lfile 1s40_aptamer_Model_1.pdbqt --nrun 100
AutoDock-GPU version: v1.6-release

Running 1 docking calculation

Cuda device:                              NVIDIA GeForce RTX 2080 Ti (#1 / 4)
Available memory on device:               10662 MB (total: 10824 MB)

CUDA Setup time 0.619260s
(Thread 6 is setting up Job #1)
Segmentation fault (core dumped)

Do you have any insights into what might be causing this segmentation fault? Could it be related to memory allocation limits or a potential code dependency on the MAX_NUM_OF_ATOMS parameter?

Thanks for your help!