4.0 Regarding the use of g(s*)t for low molecular weight proteins: The dcdt method produces a peak of g(s*) vs s* whose maximum position appears at a value of s* that is slightly less than the true value of s. The amount by which it underestimates the true value is related to the product of the molecular weight of the protein and the square of the speed and is essentially independent of the value of s. At 60,000 RPM, the shift is 1% (0.99) at 60kDa, 2% (0.98) at 35kDa, 3% (0.97) at 25kDa, 4% (0.96) at 18kDa. Moreover, the correction factors are inversely related to the square of the rotor speed. For example, a correction for a 60 kDa protein of 1.01 at 60,000 rpm would become a correction of about 1.02 at 46,000 rpm and would be 1.04 at 33,000 rpm. Below about 15-17kDa is it nearly impossible to get the boundary to clear the meniscus and so it is recommended that the time derivative method not be used below that range. This shift in g(s*) computed from the time derivative, arises from the way in which g(s*) is computed since it does not fully account for the effects of diffusion on the boundary shape. There is an exact correction for this effect, BUT it nearly completely obviates the automatic baseline subtraction that makes dcdt so useful. This shift seems a small inconvenience, unless very accurate s values are needed, for example, for shape analysis. In that case one could measure s either by the "old fashioned way" using a synthetic boundary cell, using the transport method or using second moment calculations However, given the uncertainties in the estimates of hyrdation factors, one might not find the use of the correction factors objectionable. The above set of correction factors will do well in most "everyday" situations. However, I strongly recommend that if you are interested in getting more accurate s, D and M values from sedimentation velocity analysis for proteins in the molecular weight range below 20K that you use one of the direct fitting methdos like John Philo's Svedberg program or Borries Demeler's finite element curve fitting program. Those methods work very well as long as there is no significant concentration dependence. Walter Stafford mailto:stafford@bbri.org revised: May 24th, 1999