Computing translational diffusion and sedimentation coefficients: an evaluation of experimental data and programs

Hydrodynamic characterisation of (bio)macromolecules is a well-established field. Observables linked to translational friction, such as the translational diffusion (D (t) (0) ((20,w))) and sedimentation () coefficients, are the most commonly used parameters. Both can be computed starting from high-resolution structures, with several methods available. We present here a comprehensive study of the performance of public-domain software, comparing the calculated D (t) (0) ((20,w)) and for a set of high-resolution structures (ranging in mass from 12,358 to 465,557 Da) with their critically appraised literature experimental counterparts. The methods/programs examined are AtoB, SoMo, BEST, Zeno (all implemented within the US-SOMO software suite) and HYDROPRO. Clear trends emerge: while all programs can reproduce D (t) (0) ((20,w)) on average to within +/- 5 % (range -8 to +7 %), SoMo and AtoB slightly overestimate it (average +2 and +1 %, range -2 to +7 and -4 to +5 %, respectively), and BEST and HYDROPRO underestimate it slightly more (average -3 and -4 %, range -7 to +2 and -8 to +2 %, respectively). Similar trends are observed with , but the comparison is likely affected by the necessary inclusion of the partial specific volume in the computations. The somewhat less than ideal performances could result from the hydration treatment in BEST and HYDROPRO, and the bead overlap removal in SoMo and AtoB. Interestingly, a combination of SoMo overlapping bead models followed by Zeno computation produced better results, with a 0 % average error (range -4 to +4 %). Indeed, this might become the method of choice, once computational speed considerations now favouring the 5 -grid US-SOMO AtoB approach are overcome.