ChimeraX Recipes

Speed of Biopython vs ChimeraX

Chris Moth asked why ChimeraX does not use Biopython. Biopython is a great package for doing non-interactive calculations on molecular structures. But ChimeraX is all about interactive analysis and speed is very important to make it usable.

How different are ChimeraX and Biopython in speed?

I timed opening a 2 million atom mmCIF file 3j3q and getting its atom coordinates. ChimeraX was 16 times faster reading the mmCIF, uses 3 times less memory, makes a list of atoms 100 times faster, and gets coordinates 200 times faster.

software	read mmCIF	memory used	atom list	coordinates
Biopython 1.78	131 sec	6.2 GB	0.4 sec	9.5 sec
ChimeraX 1.1	8 sec	1.7 GB	0.004 sec	0.04 sec

Timing code

Here is the code I used to time ChimeraX opening the structure

 time open 3j3q

and accessing atoms and coordinates from Python shell (menu Tools / General / Shell)

s = session.models[0]
from time import time
t0 = time() ; atoms = s.atoms ; t1 = time() ; print ('atoms', t1-t0)
t0 = time() ; xyz = atoms.coords ; t1 = time() ; print ('coords', t1-t0)

And here is the code I used to time Biopython

from Bio import PDB
parser = PDB.MMCIFParser()
from time imort time
t0 = time() ; s = parser.get_structure('3j3q', '/Users/goddard/Downloads/ChimeraX/PDB/3j3q.cif') ; t1 = time() ; print('read mmcif', t1-t0)
t0 = time() ; a = tuple(s.get_atoms()) ; t1 = time() ; print ('atoms', t1-t0)
t0 = time() ; c = [a0.get_vector() for a0 in a] ; t1 = time() ; print ('coords', t1-t0)

Tom Goddard, September 15, 2020