Tour of Learning from Docking, modelling and crystal structures

A collection of the slides shared in a combined COMP/medchem team meeting 4th June 2020.

My overview of critical compounds of the moment : Design_team_medchem_leads_1_slide_4_6_2020.pdf (105.7 KB)



Look forward to seeing the slides! And contributing in the future!


1 Like

These are my slides showing an overlay of some of the key crystal structures we are working with at the moment:
Moonshot_Struc_Overlay_040620.pdf (1.2 MB)


Some_Observations.pdf (2.7 MB)

here are my slides on the binding site from yesterday, Best wishes, Bobby


Building SAR: resolving between noise & a significant change in activity

Efficiency in building SAR can increase speed and decrease cost for drug discovery. This efficiency requires knowledge of the degree of precision (scatter or variability) in activity measurements. Underestimation of precision causes unnecessary loss of information, because some likely genuine changes are treated as noise. Conversely, overestimation of precision is misleading when noise is treated as a genuine shift.

Using 9 data sets for CVD2707 (MAT-POS-916a2c5a-2), it is estimated that at least a 3.6-fold shift in IC50 is required for 95% confidence that there is a real change in activity (minimum significant ratio, MSR) when comparing reversible inhibitors in the fluorescence assay (original format). What is a significant change in SAR.pdf (427.9 KB) This is a fairly typical MSR, but it may be changed by any modification of the assay conditions.

Similarly, the MSR in RapidFire is estimated as 3.1-fold (using 13 data sets for CVD1643 or Ebselen). This applies only to comparisons within the RapidFire assay using this format. Comparisons with fluorescence-based IC50s may show a shift in potency due to the different assay format, especially a change in the identity and concentration of the substrate.

Please contact me if you have any questions. Stay well!



Thanks @Wal, I really enjoyed reading the document – and it is a completely sensible thing to do. I’m glad to know we are within the range of acceptable MSR. One question I had is if MSR is expected to change with potency, or is expected to be constant across ranges of potency?

Thanks again for doing this analysis,

Hi Matt

Thanks for your great question. The MSR can be applied to any valid IC50 across the potency range, but only within the limits of the assay. For example, the highest concentration of test compound in the fluorescence assay is 100 uM, so any IC50 above 20 uM may be imprecise. Also, IC50 cannot be below 50% enzyme concentration, and may become inaccurate below about 3x enzyme concentration, or 15 nM. In practice, the minimum could be lower, assuming that the active dimeric enzyme dissociates into inactive monomers which do not bind the compound. Further work would be required to be confident of characterising such potent compounds – we shall all be happy when the project faces this challenge! Also, IC50 may not always been a valid measure of activity, eg if the compound is slow binding or interferes with the assay technology.

The MSRs were calculated for reversible inhibitors, so are unlikely to be valid for irreversible compounds.

All the best


1 Like

Thanks Wal,

That’s very helpful. We would be very glad to face the challenges of assaying extremely potent compounds, indeed. Hopefully, something to look forward to and hope for…


Thanks Wal for reminding everyone that statistics matter…
My question: if one had control compounds spanning the potency range with each assay run (say 4 orders of magnitude), would that allow one to confidently adjust the IC50s of all other compounds in that assay run?

1 Like

Hi Frank

Sorry I missed your question until now. I am not sure what you mean by “adjust the IC50s”. Is that to allow for the effects of changing assay conditions? If so, then the approach you suggest would give an approximate estimate if the compounds under consideration are affected the same way. All would need to be

  • rapidly reversible (slow reversibility is evident as curved time-courses). CVD2707 gives a measure of acceptable curvature for a rapidly reversible inhibitor.
  • non-tight binding (IC50 > 5x enzyme concentration), apparently OK for all compounds so far.
  • follow same mechanism with respect to substrate (the ratio [substrate]/Km can affect IC50). Most compounds are substrate competitive, so this should be OK in most cases.
  • affected in similar ways by additional factors (eg reducing agent, buffer, pH, ionic strength etc), but this is difficult to predict without doing the experiments. Accordingly, predicted shifts in IC50 need to be considered as tentative.

Hope this helps. Please contact me if I can help


Rather than considering significance of differences between pIC50 values (or ratios of IC50 values), it might be better to derive confidence intervals for ΔpIC50 values because this fits more naturally into the SAR mapping/analysis framework. One challenge when estimating precision for IC50 is that replicates are not necessarily true replicates (they may have been run one after another using the same materials).

Hi Pete

Thanks for your comments. It is difficult to derive confidence intervals for ΔpIC50 if there are less than 6 replicate determinations of pIC50 (which obviously is the case for most compounds!), because as you suggest estimation of meaningful confidence intervals requires a repeat of the whole experiment (not replicate data within a single experiment). This is due to the noise being greater between (not within) datasets (eg when comparing compounds). The MSR is based on repeat experiments with a standard reference compound and estimates the shift required for 95% confidence, please see the document above. Take care