OSM Series 4 - Next Round of Synthesis #358
Comments
|
@mattodd I understand and agree to your guiding principle of new analogs being below LogP of 3. While all of your rationale and reasoning is well thought out and sound, we might have to break with that rule for the short term. The analogs that I had proposed were based on testing a few new SAR ideas, but they also were chosen based on currently available chemicals in my inventory, feasibility of synthesis by my students, and to closely match active analogs (not necessarily metabolically stable analogs.) Essentially I chose to de-tune druglike characteristics in order to test some new SAR. If active, the goal would then be to re-tune back in more desirable physicochemical properties. Long story short, we would like to continue with our current set of targets (which exceed LogP=3) and then adjust later of they are active. Is this acceptable? |
|
If the higher logP analogues are made to address a specific question regarding SAR then I would not see any issue. |
|
Various musings: |
|
Hi @MedChemProf I'd agree with @drc007 - violations of the logP rule are fine if there's another reason to make the molecule - the Shackleford being a case in point. In any case if you want to make a molecule, then go ahead and make it! The logP "rule" is just a best guess guideline for everyone that is intended to help gross parameters of the molecule and therefore hopefully help lower clearance rates. |
|
I am very curious about the "open-ring" compounds - I predict [that] they may not be very potent because the V-like shape that I believe is crucial to the functioning of these drugs is quite distorted in the open-ring molecules - Perhaps a open-ring molecule should be created in order to test this theory. |
|
I'd need more convincing of this "V-theory" @MFernflower . Is there a correlation that shows compounds with that shape are active and those that are not that shape are inactive? We've lots of data to allow such an assessment. And let's imagine that the molecule you propose is inactive. Can we lay that inactivity at the door of the overall molecular shape, or might there be other factors? Anyone have a synthetic approach for The Triazole? I am assuming we might be able to add in TMS-acetylene via a nucleophilic displacement of a chloride, followed by deprotection, to yield the precursor alkyne. We'd then need that 4-azidopyridine. Pitfalls? |
|
Mandrake - to save people's email inbox fatigue, could you please group On 16 November 2015 at 13:07, Mandrake Fernflower notifications@github.com
MATTHEW TODD | Associate Professor THE UNIVERSITY OF SYDNEY CRICOS 00026A |
|
I understand fully, I am just leaving the idea of V-theory out in the open if someone with magnitudes more firepower than me wants to tackle it and prove/disprove it Addendum: I wish I could be more helpful than I currently am - I feel like I am parasitizing (pun intended) off of you guys and gals with my ideas.. |
|
I am happy to do similar designs as before. While updating the activity model I noticed the values of 'PfaI EC50 uMol (Mean)' for OSM-S-277, OSM-S-278 and OSM-S-279 seem to be transposed a row (compare to 'Pfal IC50 (Syngene)'). |
|
Thanks @wvanhoorn - your analysis would be very useful to this process. Thanks also for the alert about those compounds - fixed. While we work out what to do with some difficult values (those that have qualifiers, or % inhibition values at certain concs) the "mean" column is manually created. Suboptimal. |
|
Hi all,
Generally it seems that compounds need to have logP > 3 in order to have a chance at being active in the assay. Do you agree a logP > 3 rather than logP < 3 criterion is necessary to ensure compounds enter the cells and potency is measurable? As a check, here is the logP-potency correlation for series 1 compounds which shows the same logP cutoff necessary for potency.
|
|
Very interesting analysis @MJTarnowski . I guess the point is that we know we can generate potent compounds with high-ish logP values. The trick is to maintain the potency as logP is lowered. This is making the assumption that the high logP is responsible for higher-than-desired metabolic clearance rates. So the current strategy (discussed as the main item of business at the last meeting) is to make compounds with lower logP, see if they can be reasonably potent, and then assess microsomal stability. i.e. to alter the bulk properties of the compounds rather than trying to block positions on the molecules. What do you think of the strategy? |
|
As was done last summer, here is an analysis of compounds that could be made using interpolation, i.e. using core and Rgroups that have been used before. First the model was rebuilt using the latest 'PfaI EC50 uMol (Mean)' data from the master spreadsheet:
The quality of the model is sufficient to use it to select compounds. There are 104 NorthWest substituents which combined with 24 NorthEast gives 2496 possible combinations (per core, only triazolopyrazine core was used). This virtual library was filtered by AlogP <= 3.5 and no amide linkers. Two selections were made:
The selections are plotted below: Exploitation = diamond, Exploration = star, Known (already made) = square, Other suggestions from blog = triangle, Not eligible (not passing filter) = grey. The colour gradient is based on model score (blue to red = low to high).
The same compounds (minus the suggestions from the blog) are plotted in the heatmaps below. Exploitation = 1, Exploration = 2, Known = x. The top heatmap shows the effect of the AlogP/amide filter.
Structures of the selections:
Gsheets with details (including Smiles, InChi and InChiKey. I don't have a tool to calculate IUPAC names in bulk: https://docs.google.com/spreadsheets/d/1Q14mdFfLInbpqbhLcCpC-1C1pDczIc6oVLJpRZIqUQM/edit?usp=sharing Also as before some transformations have been applied to see if novel structures could be generated. All transformations are based on ChEMBL, i.e. for each transformation there is at least one pair of molecules in ChEMBL that exemplify the transformation. Only transformations that introduce a change in number of atoms <= 4 were attempted. Publicly available filter like PAINS were applied to remove unlikely products as well as QED > 0.5. As with the interpolated designs: AlogP <= 3.5 and no amide linkers. Similar Exploitation and Exploration picks were made. In the graphs below: colour gradient by model score, Diamond = exploitation, Star = exploration:
Gsheet of the above selections: |
|
I understand that what I am about to put forth is in direct contradiction to some of the analogs that my students and I are currently looking to build, but I just wanted to interject a few thoughts into the discussion. Firstly, I do concur with @mattodd concerning the need to reduce the LogP to potentially mitigate some of the metabolism that may be taking place via Cyp oxidation. While there are still a few more areas to explore some SAR, it appears to me that the group has already tried many of the strategies used to reduce LogP and introduce some added polarity. Most of those attempts has resulted in compounds of significantly lower potency. My concern is that the current core scaffold needs to retain the 3 aryl groups (the core triazolopyrazine, the aryl ring attached to the triazole, and the aryl ring at the end of the linker) in order to be active, and that the core is the primary culprit for metabolic liability. While looking at making some changes to the triazolopyrazine core, I used the predictive P450 module in Optibrium's software package 'StarDrop'. All of the analogs that I ran using their predictive algorithm resulted in one of the carbons adjacent to the pyrazine nitrogen as the potentially most metabolically liable position. It appears that this has some evidence backing it up from your earlier work looking at Metabolic ID. Below is the calculation for OSM-S-175 which had some clearance data in the master spreadsheet.
I do not know if it would be possible, but would it make sense to take one of the compounds similar to OSM-S-175, but with a significantly lower LogP and run the same MLM or HLM study? I also ran OSM-S-185 and it had a similar calculated result to to 175. I was simply trying to see even if we had a compound that met the LogP criteria, but was cleared at a similar rate, then that may indicate that the core itself is the achilles heel. Perhaps then more emphasis should be directed at a scaffold hops versus R-group modifications.
On a separate note, there was a comment on another thread describing the conformation of the various analogs as 'V-Shaped'. While thinking about triazolopyrazine scaffold hops, I have also been trying to minimize the new structures and see if they can overlap with a minimized version of some of the active compounds in Series 4. Using the simple minimization tools that I have at my disposal, most of the conformations have the aromatic ring on the linker arm distant from the aromatic ring on the triazole portion of the core. Just as a thought experiment, I docked some of the Series 4 compounds against a calcium channel that was used as a homology model to the PfATP4ase channel. A few of the docked structures showed an internal pi-stack between the two aromatic regions of the molecule that to me could be termed 'V-shaped'. I know that the site of action of the Series 4 compounds is still unknown, but it at least gave me an appreciation of a few alternative possibilities that I was not considering before. |
|
Thanks @wvanhoorn - just to clarify at the outset, the sheet that you link to at the very bottom of the post - are the compounds therein the same as the ones for which you provide structures further up, or are they distinct? @MedChemProf Yes indeed, this is an issue. Nice analysis. Interestingly the results you obtain are a little different from those from the analysis @drc007 carried out that is summarised here. |
|
@mattodd
From a physicochemical point of view, our screen selected for compounds having a larger molecular mass and a higher hydrophobicityindex than the average for the source compound collection (446 versus 385 Da; 5 versus 3.3 clogP, respectively; Fig. 2a, b). The meaning of this observation is unclear, but it may have to do with the physicochemical requirements needed to reach intracellular targets. |
|
Most interesting @MJTarnowski - is this a feature of hits (as discovered from screens) vs leads (desirable, designed) @PaulWillisMMV ? |
|
@mattodd @MedChemProf with respect to the metabolite prediction, the work I did was for the ether series not the amides, I would expect the electron-withdrawing amide to markedly change the sites of metabolism on the ring. |
|
What is bothering me is that this is technically the second time this amine did not react as expected. Several products seen in the crude NMR when previously reacted with an isocyanate. The HBTU/HOBT conditions are my usual go to for very good yields. Exam week here, so limited lab work right now. Hopefully in the next few weeks we will move more of the amine forward to try a few more conditions. |
|
Perhaps you can try something a bit stronger? PyBOP or sodium amide might do the trick (provided the amide doesn't decompose back into the parent amine during treatment or workup) |
|
Don't have any PyBOP available right now. I'll try the reagents we have in the lab first. |
|
HBTU/HOBT is certainly very reliable and usually gives clean products. You could try the acid chloride. The other possibility is that the amide is very easily hydrolysed on workup? Is it possible the amino heterocylce is a very good leaving group?
|
|
I really had no reason to suspect that the amide product would be that easily hydrolyzed, but I will keep it in mind. Barring the HBTU went bad (which I also think unlikely), the probable failure came with the amine being a weak nucleophile and a more reactive acylating agent or elevated temperatures are required. I may first try to simplify things by just taking the recovered material and attempting to acylate it with acetic anhydride. If that works and the product is stable, then I will invest in creating a hotter derivative of the more expensive 3,4-Difluorophenylacetic acid. |
|
@MedChemProf Could one treat the 3,4-DFPA with phosphorus pentoxide to create the needed anhydride?? |
|
Just an update on recent lab activity. The following three compounds were synthesized and just sent out for antimalarial testing. Two of the compounds are new and one was synthesized as a control to be able to compare backwards to results from previous screens. The control is compound OSM-W-272 (MMV639565) and its synthesis can be found here. The two new compounds are OSM-W-3 and OSM-W-4. I have entered them into the tracking spreadsheet, please let me know if someone spots any problems. The final synthetic step of OSM-W-3 can be found here and the final synthetic step of OSM-W-4 can be found here. OSM-W-3 was synthesized as part of the effort to lower the LogP and maintain activity. OSM-W-4 was one of the new scaffold hop analogs. The other Tetrazole scaffold hop analog is still in progress (and will be until September when the summer break is over.) Hopefully, we will also have a Reversed Amide synthesized by September also. |
|
Hi Chase - Fantastic. Where's the evaluation being done? In an ideal world we use similar protocols, and similar controls - @alintheopen could you please point to public domain assay details for e.g. Syngene? Or copy the essentials here. |
|
@mattodd @MedChemProf The left hand sulfur gave me an idea - If the sulfur compound proves potent I would like to see a left-hand amide: http://www.sigmaaldrich.com/catalog/product/aldrich/a1800?lang=en®ion=US |
|
Syngene Assay Details: "This protocol assesses compound efficacy against Plasmodium falciparum in-vitro. This assay is using [3H]-hypoxanthine incorporation or DNA labeling by SYBR Green as a markers of parasite growth. |
|
@mattodd @alintheopen The assay is being run at the Broad Institute in Cambridge, MA (https://www.broadinstitute.org). Since we have so few compounds, they are generously adding them into the screen that they run regularly. Below is a summary of the screening details that was sent to me: Drug assays were performed as previously described [Johnson, J.D., et al. "Assessment and Continued Validation of the Malaria SYBR Green I-Based Fluorescence Assay for Use in Malaria Drug Screening", Antimicrobial Agents and Chemotherapy, 2007, 51(6), pp. 1926–1933; doi:10.1128/AAC.01607-06], with modifications for 384-well format. Briefly, synchronized ring-stage parasites were cultured in the presence of 12-point 2-fold serial dilutions of test compounds in 40 µL of RPMI supplemented with 0.5% AlbuMAX II at 1.0% hematocrit and an initial parasitemia of 1.0% in black clear-bottom plates (Greiner Bio-one 781090). Following 72 hours' incubation under standard culture conditions, SYBR Green I dye (Invitrogen S7563) was added to a dilution of 1:5000, and plates were stored at room temperature until fluorescence signal was read on a Spectramax M5 plate reader (Molecular Devices, ex 480 nm, em 530 nM). After background subtraction and normalization, EC50 values were calculated using the Levenberg-Marquardt algorithm as implemented in the Collaborative Drug Discovery database [Hohman M, Gregory K, Chibale K, Smith PJ, Ekins S, Bunin B. Novel web-based tools combining chemistry informatics, biology and social networks for drug discovery. Drug Discov Today 2009;14:261-70.]. The primary contact that I have at the Broad published a paper on some of her screening work here: Amanda K. Lukens, Richard W. Heidebrecht Jr, Carol Mulrooney, Jennifer A. Beaudoin, Eamon Comer, Jeremy R. Duvall, Mark E. Fitzgerald, Daniela Masi, Kevin Galinsky, Christina A. Scherer, Michelle Palmer, Benito Munoz, Michael Foley, Stuart L. Schreiber, Roger C. Wiegand, and Dyann F. Wirth, "Diversity-Oriented Synthesis Probe Targets Plasmodium falciparum Cytochrome b Ubiquinone Reduction Site and Synergizes With Oxidation Site Inhibitors", J Infect Dis. (2015) 211 (7): 1097-1103. doi: 10.1093/infdis/jiu565 |
|
@mattodd @alintheopen I just received the screening results for the three compounds shown above in the thread (OSM-W-3, OSM-W-4 and the control OSM-W-272 aka MMV639565). Using the screening protocol (also listed above in the thread), the compounds were screened against two malarial cell lines - 3D7, a generally drug-sensitive strain; and Dd2, a multi-drug resistant strain (Choroquine and folate biosynthesis inhibitors). I have attached a captured image that shows the EC50 curves because for some reason Github was not accepting the pdf upload. |
|
Thanks Chase! Some really interesting results here, particularly with respect to the para sulfone - definite food for thought. Target ID work also sounds interesting and will be excellent to compare this data to the other Scaffold Hop analogues. Nice work! Alice 👍 P.s I think that images have to be png format for GitHub |
|
Anyone want to make a left-hand para-methoxy? Dont think left hand electron donor was ever tested? |
|
@mattodd Using DataWarrior to run the calculations, OSM-W-272 has a cLogP=3.79 and OSM-W-3 has a cLogP=2.08. The reduced potency aside, is this close enough of a comparison to be used in a MLM / HLM study to determine if the strategy of reducing the LogP is a viable way to attenuate metabolism? |
|
Hi @MedChemProf - it's probably worth a shot, yes. Do you have a calculated logD - can't remember if DW provides that. (Cross-link to #388 for reference). But we'd probably only do this if the frontrunner set in #400 didn't provide us with enough data on logP/logD vs microsomal stability, I'd guess. It would need to be worth our while you shipping compounds to wherever the frontrunners end up (likely Monash, but not yet confirmed). |
|
@mattodd Should we close this thread as you suggested and begin new ones on any series we are still working on? Thanks. |
|
It's high time we closed this, but the reason it's not happened yet is that we a) need to ensure that any compounds that have been biologically evaluated have been added to the wiki, and b) that any useful suggestions in this thread have been captured somewhere, which goes to #440. Related to a), Chase, do you have a chemdraw file that we could use to generate pictures to add to the wiki? i.e. structures, codes and potencies, along with the URL for the relevant biological data? |
|
I do have a Chemdraw file with the above compounds. GitHub does not accept chemdraw file types, where is the file going? |
|
Dropbox might work
|
|
Yes, this has been a perpetual issue - a shared file store, and which to use? A non-trivial problem. I suspect Dropbox would work, as, perhaps, would Slack. I had hoped we might be able to use Github for this, given that we're all already here. The wiki images are helpful for summarising progress to date. The original intention, 5 years ago or so, was that the writing of the wiki would be essentially the same activity as writing the paper, to avoid duplication. Results would be folded in as they came in. I don't think there's a platform for this, and the problem with the wiki is that it's another sign in. However, perhaps people could look at the repository I started for the first Series 4 paper which is something we need to re-activate in the coming months. There is a Word document there that can be downloaded, edited and re-uploaded by, I think, anyone in the associated team. One can post chemdraw files there, and, I think, lots of other filetypes. Perhaps it's time for us to step outside of Github as merely a To-Do list and start using its filesharing capabilities more properly for paper writing. I wonder, then, if there's a way we can maintain a Word file as the current status of the project document, with this To Do list acting as a place to discuss things related to the paper specifically. Alternatively the paper is constructed in these windows with markdown, and then converted to a Word document later. I suspect that is a better option, though slightly beyond my technical abilities. Writing the paper while maintaining project status involves a complex web of activities that needs to make sense to everyone, so I'd be grateful for suggestions as how best to do this (in a way that permits open filesharing). Obviously the size of this question may require me opening a new issue. In fact, let me do that now. Hold on. |
|
NB update related to mention of Lilly's OIDD above, the platform is closing in Q1 2019. #341 |
It’s decision time on the next set of targets.
Alice just mailed the latest set of analogs she has shepherded through (OSM-S-291 to OSM-S-301). This set contains a few compounds that will add to the data surrounding tolerance in the northeast (arising from our 1st year undergrads and Seb Dath’s work (1, 2), and some compounds probing tolerance of a benzyl-linked pyridine in the northwest - we could still explore benzyl substituents more, particularly with blocking groups (as per Chris’s suggestion) - see below for more on this.
But which molecules are next? Generally we need to lower predicted logP, work towards lowering metabolic clearance times while maintaining or improving potency.
Analog design in this series has previously been discussed: June 2015 ELN post, March 2015 ELN Post, December 2014 Xmas Shortlist, and in issue #301.
The most recent biological evaluations were:
June 2015
May 2015 Edinburgh
November 2014
July 2014
and the most recent online meeting was: July 2015
Summarising the data, and all the many and various community suggestions to date (particularly from the last meeting), we can arrive at some guiding principles:
Benzylic blocking groups have not to date resulted in significant improvements in metabolic stability, leading to the suggestion that improvements in gross parameters of the molecules should be improved, most notably logP. cLogP for any proposed analog must now be below 3. This rules out several interesting compounds (red crosses, below) that need to be redesigned, and places a question mark over marginal compounds (orange crosses). For example those suggested by @mrinalkundu would need to be modified with this design criterion in mind - the same goes for the updated set from Mrinal. We do still need to show better the correlation between measured and calculated solubilities, though. There is no clear trend between logD and metabolic stability, for the limited number of compounds we have measured (Trends in logD vs metabolic stability? #333). Note that the cLogP values are only approximate and are in Chemdraw sensitive to whether X-H bonds are drawn out explicitly.
However, blocking groups next to oxygens in the northwest linker should be explored, based on the metabolic ID work discussed in the last meeting, since it could be that O-dealkylation is occurring next to the TP ring. See also Metabolism #334
We should probably avoid unsubstituted pyridines, generally, though the Sydney Grammar samples below should be taken through to completion given that we have them!
To generate an IVIVC, Is there an In Vitro - In Vivo Correlation for Series 4? #351, it might be necessary to re-synthesize specific compounds, but we await input on this point from David Shackleford.
These considerations lead to three groups of compounds - the ones that ought to be made immediately, those that are highly desirable and those that are more speculatively interesting.
The Immediates
Nemesis: Alice is in the process of making this, though it’s being a tough nut.
The CRO: a compound we’ll purchase as a test of the receipt of open data, as described in this background post and in #292.
The NW Aliphatic: The des-methyl alcohol is being prepared by Alice.
Sydney Grammar Pyridines: These are still being purified - Paul King was looking at them. We’re at the stage of attaching the nucleophile in the last step.
The Dimethyl Amine: Needs to be verified based on data from the last round since methylation of that amine appears to have a profound effect on potency.
The Homologous Alcohol: Good to have, but too lipophilic so needs a re-think.
The Homologous Amine: Ditto
The Shackleford: Blocking group next to alcohol, as per last meeting. This compound fails on cLogP, but may be worth making anyway to test clearance rate. Related issue was to search for commercially-available aryl ethanols #302.
The Swain: Blocking group and benzyl ether. Ditto - may be worth it despite the poor cLogP, or structure could be tweaked? Benzyl ethers still badly explored in this series, so the 2-, 3-, and 4-Cl substituted phenethyl analogs could be checked (suggested by Chris) and then drawn by Alice.
The Desirables
Reversed Amides: These have been discussed, e.g. here. The synthetic route to this reversed amide series could be through simple amination of the common chlorotriazolopyrazine intermediate, which both Alice (e.g. here) and Chase are interested in. Chase also suggested azide displacement of Cl followed by reduction to the amine.
The AZ Idea: A group of young scientists at AstraZeneca that includes former OSM-er Matt Tarnowski recently suggested a cyclic analog shown below, and our young contributor Mandrake also suggested some ideas around that molecule. Some cyclic groups have indeed been investigated in this position generally with low potencies - see the summary below. The group joining the side chain to the TP ring (red arrow) can’t be divalent oxygen; we know simple amine linkers are poor choices (e.g. OSM-S-190), yet amines have typically been the attachment point for rings (see below figure). Carbon linkers have occasionally been OK (e.g. MMV 669304, MMV670243). Perhaps a more polar cyclic group at this position would be OK with a carbon-based linker provided the northwest ring is positioned well. Metabolic clearance of carbon-linked compound MMV669304 was atrocious (MMV669304 below), while the related cyclic amine linked MMV668957 was much better (yet much less potent).
The Hypotheticals
Crazy Hemiacetal: Hemiacetals and hemianimals were suggested in the last analog analysis round. They are expected to be too labile, even though acetals themselves can be OK. While exhibiting a blocking group and good predicted hydrophobicity, this compound is probably too risky.
Other Linkers: Some of the SAR data suggests the linker between pyridine and Ar ring in the Northwest doesn’t really matter, leading to suggestions of ketone, sulfoxide and sulfone linkers from Chris and a urea suggested by Chase. We could also investigate a triazole linker (using an alkyne on the pyrazine to avoid a nitrogen-based linker atom on the ring) and an alpha-trifluoromethyl amine as an amide isostere, originally suggested by Chris Burns.
The Tianyi: Rationale for this molecule is that the imidazopyrazine MMV669846 retained potency (vs parent MMV639565) with only a slight deterioration in clearance time (see below), suggesting it might be worth re-examining this compound, but with a blocking group on the imidazole (methyl shown). Named after the student who started looking at this compound, Tianyi Zheng.
Chase Scaffold Hops: suggested here. These will require development of new methods if they’re deemed to be good swaps. Any opinions?
Misc Other Points
Now that we have more data in the Master Sheet, I wonder if we might be able to ask you, @wvanhoorn , to re-do your analysis, prioritizing compounds with untried motifs, using data from Series 4, and with an ALogP cutoff of 3. There was also mention of results from a matched pair analysis by Jeremy Besnard - I think the structures are here. It’d be interesting to see what you get now.
We will shortly be engaging with the OIDD automated synthesis group, and that some of the above synthesis could be part of that collaboration if, for example, a range of reaction conditions need to be screened or a number of analogs need to be generated. Volunteer Request: Submitting Proposal to Lilly OIDD Synthesis #341
How to Comment on These Suggestions
Please do comment/suggest/criticise. Please do make any of these compounds, provided you share the data so we don't duplicate.
Commenting can be done below, or on Twitter or G+ if you have accounts there. If you have to use email (please don’t) then you can do that.
Suggesting structures: you can paste SMILES, or you can easily drag and drop pictures into comments on Github now.
Deadline
Next shipment of compounds to Syngene: December 15th.
(cheminformatic strings for all molecules mentioned in this post are over on the OSM ELN here)
The text was updated successfully, but these errors were encountered: