Extrapolation beyond limits

Parameters with steep slope due to lack of diverse training data tend to produce unphysical k values when extrapolated below 1.

An edge case of t72 being replaced by an interpolated parameter TIG8 that has a steep slope.

t72 smarts: "[#8X1:1]~[#7X3:2]~[#6X3:3]~[*:4]"

t72: k = 0.74 kcal/mol

TIG8: k1_bo1 = 1.71, k1_bo2 = 10.5 kcal/mol

this results in the k value going to zero at bondorder of 0.8, whereas this may not be true.

 

 

In the new amide dataset created by Simon Boothroyd, molecules that match t72 have a WBO spanning the range of 0.7 to 0.9

Histogram of wbo distribution for this parameter

Few molecules with a lower WBO shown below

ELF10 WBO

Chemical structure

 

ELF10 WBO

Chemical structure

 

0.75

 

0.75

 

0.79

 

0.72