dpocket and fpocket return different drug scores

[radakb]

Describe the bug
fpocket and dpocket compute different drug scores for otherwise identical pockets (all other descriptors appear to be the same)

To Reproduce
Steps to reproduce the behavior:

1. Take any holo pdb and create an analogous "apo" pdb by deleting the ligand
2. Run fpocket -f apo.pdb -d | head -n2 > fpout_fpocketp.txt and echo "holo.pdb <residue name>" > input.txt; dpocket -f input.txt
3. Compare descriptors in fpout_fpocketp.txt and dpout_fpocketp.txt:

These are in different formats and differ in contents so I ran the following script to compare a few obvious ones.

for val in drug_score hydrophobicity_score volume_score polarity_score charge_score mean_loc_hyd_dens as_density
do
        fpock=`awk '{if(NR==1){for(i=1; i<=NF; i++){ if($i == "'$val'"){ j=i; }}}else{print $j}}' fpout_fpocketp.txt`
        dpock=`awk '{if(NR==1){for(i=1; i<=NF; i++){ if($i == "'$val'"){ j=i; }}}else{print $j}}' dpout_fpocketp.txt`
        printf "%20s % 8f %8f\n" $val $fpock $dpock
done

Expected behavior
Unless I am misunderstanding something about how dpocket operates, I would have thought all descriptors would be the same.

Screenshots
This probably wouldn't help, but I visualized and diff`ed the vertices pqr files and everything is the same except for the header.

Desktop (please complete the following information):

• OS: Ubuntu 24.04
• Fpocket 4.0

Additional context
My original goal was to make a quick workflow to report the drug score for the pocket most overlapping with a given holo structure -- this finding doesn't hinder that goal because I will just give deference to fpocket and choose the pocket with the highest overlap (not hard to script), but I assumed dpocket already implemented this.

[radakb]

Update: Digging into the code I'm seeing that the same descriptors have different names, specifically "surf_pol_vdw22" in dpocket vs "af_pol" in fpocket -- these quantities are clearly different in the outputs noted above and go directly into the drug score. "as_max_dst" is only outputted by dpocket.

Maybe a better question is "are these quantities supposed to be different?"

[radakb]

Still digging in the code. There seems to be a difference in the fpocket vs dpocket calculation in the asa.c::set_ASA() code path. The results for nnburried differ at various points and this leads to different results for surf_vdw-type descriptors. The number of vertices are the same but the area contributions per vertex are not.

I'm happy to help track this down and supply a fix (assuming there is a bug), but any expert advice on the algorithm would be dearly appreciated.