The example input files in this directory calculate the binding of a small molecule (acetazolamide) to a medium-sized protein (human carbonic anhydrase).

The UHBD calculations where performed using a van der Waals surface definition for the dielectric coefficient. This is simulated in the APBS input files by setting srad to 0.0.

Input File	Description	APBS Version	Results (kJ/mol)	UHBD (kJ/mol)
apbs-mol.in	2-level focusing to 0.225 A, VdW surface, srfm mol	3.0	-52.4648	-70.00
		1.5	-52.4648
		1.4.2	-52.4648
		1.4.1	-52.4648⁶
		1.4	-51.4648⁵
		1.3	-52.4647
		1.2.1	-52.4647
		1.2	-52.4647⁴
		1.1.0	-52.4669
		1.0.0	-52.4669
		0.5.1	-52.4669³
		0.5.0	-52.1062²
		0.4.0	-52.4414
apbs-smol.in	2-level focusing to 0.225 A, VdW surface, srfm smol	3.0	-54.0598	-70.00
		1.5	-54.0598
		1.4.2	-54.0598
		1.4.1	-54.0598
		1.3	-54.0598
		1.2.1	-54.0598
		1.2	-54.0598⁴
		1.1.0	-54.0587
		1.0.0	-54.0587
		0.5.1	-54.0587³
		0.5.0	-54.7039²
		0.4.0	-54.0393¹
		0.3.2	-57.1192
		0.3.1	-57.1192
		0.3.0	-57.1192
		0.2.6	-57.1192
		0.2.5	-57.1192
		0.2.4	-57.1192
		0.2.3	-57.1123
		0.2.2	-57.1123
		0.2.1	-57.112
		0.2.0	-57.711
		0.1.8	-58.51

¹ The discrepancy in values between versions 0.4.0 and 0.3.2 is most likely due to three factors:

A bug fix in Vacc_molAcc which removed spurious regions of high internal dielectric values
A switch in the algorithm used to compute the dielectric smoothing for srfm smol
The addition of the Vacc sphere density (sdens keyword) as a variable and a change in the default sdens value from 3.0 to 10.0

² The discrepancy in values between versions 0.5.0 and 0.4.0 is most likely due to the following factor(s):

A change in the autofocusing routine for APBS

³ The discrepancy in values between versions 0.5.1 and 0.5.0 is most likely due to the following factor(s):

Bug fix regarding multipole behavior for neutral proteins

⁴ APBS 1.2 has switched the multigrid smoothing algorithm from standard Gauss-Seidel to Gauss-Seidel red/black in order to facilitate parallelization. This switch has caused small differences in individual calculation energies which, when combined to the final answer, create larger errors (up to 0.04%). These errors can be reduced by resetting the APBS error tolerance to 1e-9 or smaller values. For a more detailed explanation, please see the APBS FAQ, [here](http://www.poissonboltzmann.org/docs/apbs-faq/#sources error calculation).

⁵ The discrepancy in values between versions 1.3 and 1.4 is most likely due to the following factor(s):

Translation of contrib/pmgZ library from FORTRAN to C
Differences in numerical implementations between FORTRAN and C compilers result in small round-off discrepencies
Small margins due to these round-off discrepencies acumulate in the computations

⁶ The discrepancy in the result between versions 1.4 and 1.4.1-binary is most likely due to a reporting error.

Please see the ChangeLog or the APBS website for more information.