All Notebooks | Help | Support | About
23rd July 2015 @ 01:56

Meeting recording. Slide decks are attached. One used is v8.

Slides 2–5 (2:00 to 7:00). Potency measurements on the various batches of new Series 4 analogs.

July 2014: OK potency, poor cLogP. Around this time hERG data had led to a deprioritisation of the amide sub-series in favor of the ether sub-series.
Nov 2014: ca. 200 nM potencies with OK cLogP values. Pyridine analogs found to be non-potent.
May 2015: Edinburgh data. Triazolopyridines not potent.
June 2015: Better cLogP values and bolder structural changes, but unfortunately not potent.

Chris Swain asked whether reverse amides were made. No, though there was a reverse sulfonamide (MMV669103) and a transposed reverse amide MMV672990.

Slides 6–9 (7:00 to 23:00). Analysis of Metabolic Stability/ID Data (Master Post on known data) (wiki section)

Generally: the in vitro data for several analogs are not good enough, but are not too bad.

Slide 6: Experimental data suggests oxygenation is occurring, but it is not clear where on the molecules.

Slides 7–9: Implication is that use of blocking groups, e.g. in the benzylic position, is not helping a great deal. It is not predicted by either the aldehyde oxidase data or Chris Swain's predictive modeling that the TP ring is a problem. So what is the problem? Perhaps improvements in the gross parameters of the molecules should be the priority?

David Shackleford (Monash) (Audio level low): O-dealkylation could be the real problem. If the ether series are cleaved where indicated on slide 6, either of the two fragments could itself be oxygenated (i.e. a secondary metabolite). A possible test would therefore be to screen the two fragments, derived synthetically.

If O-dealkylation is a problem, then blocking groups next to the oxygen would become a synthetic priority. This could be included anyway as a strategy. Leading to: Next analog set planning: GHIxxx.

Paul Willis: Before doing detailed analysis on fragments, we should establish whether the in vitro data are well correlated with, and hence predictive of, in vivo data (i.e. an IVIVC). David: Let's check the existing data, and are any data missing? We could take compounds with low, medium and high in vitro stability into in vivo to check this correlation. Action: GHI351.

Chat activity relevant to this section:

Chris Southan: Do you do LC/MS for these experiments? (Ans: can check raw data files linked here)
Chris Swain: Remember O-dealkylation would leave Oxygen on triazolopyridine ring: N1C=C(N(C(=N2)C(=CC=C3C)C=C3)C(C=1)=N2)O . Structures in GHI334. (Ans: No, the cleavage is the other side of the oxygen atom)
Paul MMV: Chris Have you run the full set of compounds through your predictive Clint programme / doe sit get the rank order right, how validated is it to use in target prioritisation.  Also does it look at species differences, in some cases cpds look better in man  

Chris Swain: I've only looked at representative compounds, and the program only predicts potential sites of metabolism not rates.

Paul MMV: OK I got confused by the slides and thought some of the the EH were predictions

David Shackleford: Paul - I suspect that the EH values are predicted from the Monash experimental CLint values (Ans: Yes, these are predicted based on experimental values)

Actions: Mat Todd to share all data with David Shackleford and evaluate whether we have in vitro-in vivo correlation. If not, should we submit low-medium-high clearance rate compounds to start to generate an IVIVC. Could later submit the O-dealkylation fragments for follow-up analysis to see if the oxygenated products derive from them. Done: GHI351.

Slide 10 (23:30-26:45) LogD vs. clearance data. No obvious trends: GHI333.

David Shackleford: important to consider logD vs. Volume of Distribution (VoD). Clearance impacts oral bioavailability, but VoD impacts on half life. For any compound where we have in vivo PK, particularly after IV administration, we should be looking at both clearance and VoD in the context of the physchem parameters, whether measured or in silico. We should try to get VoD data (measured/predicted) for compounds for which we have in vivo data. It would be useful if the calculated VoD values were predictive of the measured values - that would save a lot of time. GHIXXX.

It would also be interesting to see how well the measured solubilities (logD) are predicted by common software (cLogP): Done: GHI333.

Slide 11 (27:16-29:00) Can we modify triazole in MMV639565? Both imidazole compounds MMV669846 and MMV670250 have been examined. Tianyi Zheng was due to resynthesise the imidazole system in MMV669846, but the project ultimately turned more towards trying to find new synthetic methods for functionalising the triazolopyrazone core through bromination. Imidazole analog remains of possible interest, particularly if the C-H is substituted.

Slide 12 (29:00-29:45) Pyrazine modifications. Poorly tolerant to substitution. The northwest 6-position has not been explored. Tolerance there would permit an attachment point for pulldown experiments.

Slides 13-15 (29:45-34:12 with temporary SNAFU in period 33:00-34:00 where video froze and may be skipped)) Possible/residual targets suggested by the community.

Action: resolve GHIXXX (purchase from CRO) either way.

Slides 17-25 (34:12-46:44) Outline by Willem van Hoorn of approach used by Ex Scientia to predict new analogs worthy of synthesis, recently applied to Series 4.

Contrast between whether we want an exploitative approach (highest model scores, closest to known actives) or an explorative approach (go after chemical space that has not been well explored, meaning one expects fewer actives but there is likely to be more novelty). Matched molecular pair analysis can be deployed. Willem wondered whether you can use data from all the series?

Action: Re-run the analysis now that we have more data in the Master Sheet. GHIXXX.


Chris Southan: Willem - any 3D aspects in the modeling ?

Willem van Hoorn: No it's all based on 'flat molecule' fingerprints


Slide 26 (46:50-47:32) Suggested composite structures from Mrinal Kundu.

Slide 27 (47:33-48:34) Suggested alpha-CF3 amine as isostere for amide, from Chris Burns. Lit sources: 1, 2. Would be a good project for someone to try. Isostere may remove problem of amide side chain in Series 4 which is electron withdrawing, potentially making compounds better substrates for aldehyde oxidase. Write up as synthetic branch? GHIXXX.

Slide 28 (48:35-53:55): Debrief on data management by Chris Swain. The Google Master Sheet of molecules now allows parallel entry by anyone, and downloading of the whole dataset. Use of Luc Patiny's Cheminfo to view molecules. Can also use proprietary software Vortex to view the molecules. Chris wrote an iPython notebook to allow similar viewing with open source tool, using RDKit to calculate properties. These approaches use the live dataset on the fly, which is part of the power.

Slide 37 (54:00, actually starts 55:44 ends 58:45): Chris Southan. Compounds are imported into Pubchem when a deposition is made to ChEMBL, but that leaves out the molecules in the project that have not yet been so deposited. If we are to capture the data more continuously into Pubchem, how do we do that? We also need to be careful of the number of synonyms each molecule possesses: how? Separate data management meeting to discuss: Action: Assemble agenda/time for this: GHI XXX.

(57:00): Action: Devise initial list of compounds for next stage of synthesis, then potential follow-up meeting: Existing locations for this: Post and GHI301. Now: GHIXXX.


The full content of the chat window is contained in the attached file.

Original links:




Post authored by Mat Todd

Attached Files