Meeting Discussion Point 1: Solubility

[mattodd]

Background to Meeting on May 24th 2016 (#386).

Question: Are we right to target improved aqueous solubility as our key strategy?
Question: Might improvements in aqueous solubility also improve (slow down) metabolic clearance?
Question: What are the best strategies for improving the solubility of new Series 4 analogs?
Question: Is the informal cLogP cutoff value of 3.5 for new analogs in this series wise?

Resources:

OSM Series 4 summary wiki
Lab books: Labtrove, Labarchives (Ed Tse, Chase Smith).
Compound Database: All OSM Compounds in Google Sheet, interactive view on Cheminfo, direct download of Series 4 SDF.

As discussed at #333, we have measured logD data and metabolic clearance data for 8 compounds in Series 4, from Monash. The data are all contained in the main sheet, but the relevant data are also in this daughter sheet for convenience. @MJTarnowski 's analysis suggested a logD - clearance correlation. @drc007 showed that Chemaxon gave good predicted logD values.

Comments/suggestions below are welcome, as well as (briefly please) during the webinar meeting and its follow-up (#394).

Relevant background:

1. Many of the most potent Series 4 compounds have cLogP values of > 4, e.g. MMV639565, below.
2. What is known of metabolic clearance is on the wiki, summarised in this post, and was discussed at the last online meeting. It seems as though traditional blocking strategies do not lead to significant improvements in metabolic parameters.
3. A solution might be to increase aqueous solubility, invoking the assumption that this will lead to reduced clearance by CYP-related processes.
4. We have informally adopted a maximum cLogP of 3.5 for new analogs in this Series, unless the compound is needed to answer some other question.
5. Addition of polar groups to the northwest and northeast have resulted in potent compounds with better predicted solubility (e.g. alcohols in northwest sidechains in compounds MMV672687, MMV672723, MMV670947 (below)), but other strategies along these lines have obliterated potency (e.g. MMV670936, though unsubstituted pyridine in northeast is bad).
6. The amide sub-series of Series 4 was de-prioritised when these compounds appeared to show increased susceptibility to aldehyde oxidase activity, probably arising from the electron-deficient triazolopyrazine ring. The reversed amide or Burns analogs (Meeting Discussion Point 3: Potency #390) might be appropriate isosteres should we decide to return to such compounds.
7. Current compounds being made in this space include cubane analogs and oxidised sulfur side chains (Meeting Discussion Point 3: Potency #390) to attempt to deplanarise and/or increase solubility.

Question remains: do people have suggestions for more polar groups that could be included (with literature justification for such targets being sound suggestions).

[MedChemProf]

@mattodd I think this strategy has a lot of merit and is a reasonable approach to reducing the affinity of the compounds to one of the CYP's (if that is in fact what is the major metabolizer of the Series 4 compounds. I know it has been suggested otherwise.) We are still struggling with the synthesis of the reversed amides, so I am not sure we would have the bandwidth to make, but might I suggest again the synthesis of the bis-sulfone analogs shown below:

Obviously we would have to confirm activity to be truly useful, but the idea was based on similar antimalarial analogs in the literature (J. Med. Chem (2015) 58, pp. 8713-8722 paper on some antimalarial imidazopyridazines and pyrazolopyridines.) These analogs have a LogP much lower than the 3.5 cutoff. Aldrich does sell the para-substituted methylsulfone aldehyde for the right hand portion of the analogs. The left hand meta-methylsulfone benzyl alcohol and ethanol pieces are also available from Aldrich as the aldehyde and carboxylic acid respectively, so each would need to be reduced.
Alternatively, a hybrid could be synthesized that the LogP is below 3, but not nearly as low as the above two:

Finally, I think the notion of keeping the LogP lower (3.5 cut-off) is a good idea, but with so much variability on different LogP calculators, I would be hesitant to throw anything out that is testing a new idea. I certainly am not proposing making compounds for the sake of numbers.

[MFernflower]

@MedChemProf @mattodd @drc007 What about doing something like this?

Or even this:

[drc007]

@MedChemProf @MFernflower @mattodd As a medicinal chemist who has an interest in cheminformatics I think I should add a note of caution regarding calculated properties, especially solubility. If you are suggesting modifications it would be better if you could find experimental evidence to support it. I've written a page of aromatic bioisosteres (http://www.cambridgemedchemconsulting.com/resources/bioisoteres/aromatic_bioisosteres.html) one that I'd highlight is pyridazine for which there are several examples where it has been used in drug-like molecules (http://pubs.rsc.org/en/Content/ArticleLanding/2011/MD/c1md00074h#!divAbstract) to reduce LogP and improve solubility.

[MFernflower]

@mattodd @drc007 A right-hand pyridazine does seem like an interesting thing to try -
is metabolic stability a concern when using this ring however?

Adendum: There also exists a benzene bioisostere for the wicked: http://blogs.sciencemag.org/pipeline/archives/2007/11/02/one_for_the_brave

[drc007]

I've not seen anything to suggest metabolic stability is an issue, do you have a reference?

[drc007]

Well Pfizer recently bought Anacor Pharmaceuticals Inc. in a deal worth more than $5 billion, their area of expertise is boron containing compounds.

[MFernflower]

@mattodd I Just came up with a hybrid molecule that combines Chris's right-hand pyridazine (to my surprise these rings are very stable) and Chase's left-hand arylsulfone

http://www.chemicalize.org/structure/#!mol=O%3DS%28%3DO%29%28C%3D1C%3DCC%28%3DCC%3D1%29COC4%3DCN%3DCC3%3DNN%3DC%28C%3D2N%3DNC%28OC%28F%29F%29%3DCC%3D2%29N34%29C&source=calculate

I managed to get the predicted logp down to 0.65! (the lowest predicted for any s4 compound???!??)

[mattodd]

Roger Bonnert (formerly an Associate Director and Project Leader at AZ) emailed after attending the meeting (reproducing with permission): "With solubility I think you are looking at correlated parameters in that you would expect log D to correlate with solubility and met Cl. So its reasonable to use sol (also a desirable property) as a surrogate for Cl (or just use log D). I assume these are not crystalline solubilities and ultimately that will be important - it might be worth getting some crystalline solubility data if you don't already have as things may be worse. I have found adding a chiral centre often improves solubility particularly from crystalline material."

Thank you Roger - it certainly helps planning if we can broadly assume such a correlation, while understanding that there may be outliers. By "crystalline solubility data" do you mean "kinetic" solubility, which is typically was is measured by default by Monash I think - e.g. the most recent solubility data we have. Or do you mean measuring the same way with compounds we have demonstrated to be crystalline - not something I think we've ever looked at here. (Roger replied just before I posted this: "It's just measuring from a crystalline form. It doesn't need to be a complex 'spinning disk' type solution rate that you would do in development"). This is certainly something we can do when we scale up synthesis of frontrunners, so we're clear we have something crystalline (which we should discuss in #394).

Chiral centres - this is certainly a feature of several of the proposed compounds (#390), and of the series frontrunners. Interesting that there is such an effect, which presumably arises from stereogenic centres (typically) making a molecule less "planar" to a first approximation, vs. methylenes - i.e. the effect is nothing to do with the chiral compound's enantiomeric excess? (Again, Roger replied by email just now "I agree with what you say about the chiral centre. In crystals I think the centre is disrupting the crystal structure and packing so you can see some good improvements.")

[mattodd]

Charlie Mowbray, Head of Drug Discovery at DNDi, emailed after attending the meeting to say (posting with permission) "I think that reducing logD to increase solubility may also improve metabolic stability. I do not think that improved solubility per se will improve stability. Less lipophilic compounds will partition less well into hepatocytes and this helps reduce metabolism."

Thanks Charlie. So our focus ought to be logD rather than logP? This is something that has come up before, and is I think something we need to take on board. Chris Swain has shown (#333) Chemaxon is pretty good at logD prediction.

Charlie replied just as I was posting: "I wasn’t focussing on logD being more important than logP (they can both tell us useful things), more on reducing lipophilicity (rather than increasing solubility) to reduce the rate of hepatic metabolism."

[mattodd]

David Hong, Chemistry Team Leader in the A•WOL Project in the Chemistry Department, the University of Liverpool and Liverpool School of Tropical Medicine, emailed after attending the meeting, to say (posting with permission) "You are linking metabolic stability with aq. solubility, but should you more fundamental link both metabolic stability AND aq. solubility with Log D or cLogP?"

I replied: Thanks David. I'm not sure I understand. Could you elaborate a little? The relevant data on 8 compounds are here, and Matt Tarnowski's analysis is here.

David kindly clarified by email: "To answer the question in your reply post, although aq. solubility (also solubility in general) can be affect by other factors i.e. π-π stacking etc., in general, it is closely correlated to LogD/cLogP, especially for a group of compounds with similar structures. However, to confirm this correlation, you need to have reasonable number of data points (n=8 is not quite enough statistically), and ideally, all the data should be generated by the same assay (or at least using the same method). From my experience, even aq. solubility values can vary (sometime quite significantly) from measurements done in different labs or by different assays. It may not have anything to do with the validity of assays, but in reality the measuring conditions can change the final readout to a certain degree. If you use data generated by different lab or different assay, it might affect your analysis. It’s a similar story for the correlation between liver microsomal stability and LogD/cLogP. Although there are other factors can affect the liver microsomal stability, in general, it should also track with compounds’ lipophilicity (LogD). However, as liver microsome turnover assay is a biological assay, the data (either t1/2 or CLint) can be very different between different assays for the same compounds. So if you try to analyse the data, you may need to bear this in mind regarding how comparable between data generated by different assays. I feel this is an inherited issue for an open source drug discovery project, and you may need to consider some more innovative way for more robust analysis of biological data."

Thanks for this, David. Yes, we're always conscious of this issue of inter-assay variability (it's known to many of us, and dates back a few years within OSM) and is an inherent issue for a distributed project. Controls can be used to allay major concerns in potency assays. In fact the 8 compounds referred to were indeed all analysed in the same assay in the same lab (MIPS at Monash Uni) and we would use the same facility again in further analysis of more compounds. Overall we would as a consortium try to minimise variation, but would use distributed labs where that was seen as non-critical.

[MFernflower]

@mattodd

If ionization is important for potency and metabolic stability - Could something like this prove very potent?
http://www.chemicalize.org/structure/#!mol=FC1%3DCC%3DC%28C%3DC1%29CCOC3%3DCN%3DCC2%3DNN%3DC%28N23%29N5CCN%28C%3D4C%3DCC%28%3DCC%3D4%29Cl%29CC5&source=calculate

The other idea I had was to somehow integrate a quaternary ammonium group ( perhaps something a bit like this: https://t.co/v284ZlGH4j ) into one of our drug molecules and see how potent that makes it (possible off target of nACHR may be a problem in vivo but should not effect in vitro assays)

@MedChemProf

[mattodd]

David Hong followed up by email to me with an interesting comment:
"Ultimately, you may need to consider if all these physiochemical and DMPK properties are really critical to your final goal which is to identify an antimalarial drug candidate. There are some marketed antimalarial drugs have relatively poor aq. solubility, i.e. Atovaquone or relatively low metabolic stability, i.e. Artemisinins, but they haven’t stopped them to be good antimalarial drugs. I feel to make an antimalarial project with true potential, potency (driven by interaction with biological targets, not just lumping up lipophilicity), efficacy and novel mode of action (or robust understanding of mechanism) should be the main drivers in the programme."
Thanks again, David. Certainly that is my view, where we're using physicochem/DMPK as guidelines. We should always be vigilant to spot outliers with potential.

[MFernflower]

@mattodd I found a group offerin|g plasma binding assays and other ADME screening tasks that is based in the U.K Perhaps we ask them to do the assay on some of our molecules Pro Bono? (im really curious about how the cubanes and the thio series will bind to plasma albumin) http://www.cyprotex.com/admepk/protein_binding/plasma-protein-binding/

[mattodd]

Am assembling files for Wednesday meeting (#394). Here are some of the suggestions above for improving solubility, sketched out. Let me know of any changes/additions but we can discuss in meeting. Chemdraw predicts cLogP 3.6 for the interesting azaborine structure shown (may not be the best azaborine analog, synthetically) which seems too high to me. DataWarrior gives cLogP 1.7, but refuses to calculate logD for the compound.

[MFernflower]

@mattodd great graphics as always. What do you think of my idea to combine the LH sulfone with a para DFM pyridazine on the RH side?

http://www.chemicalize.org/structure/#!mol=O%3DS%28%3DO%29%28C%3D1C%3DCC%28%3DCC%3D1%29COC4%3DCN%3DCC3%3DNN%3DC%28C%3D2N%3DNC%28OC%28F%29F%29%3DCC%3D2%29N34%29C&source=calculate

Also, while the azaborine seems like a really fun thing to try - it's synthesis would most likely need to be directed to a CRO or custom order synthesis company as the process for forming the hetrocycle involves HF and some very toxic metal carbonyls

@drc007

[mattodd]

As per frontrunner discussion at #400, desired solubility of compounds for late lead status is said to be 10 to 100 micromolar. I can't find this figure on the MMV website (anyone?) but it's in this article. If we assume a MW for a series 4 compound of 400 then this = 40 micrograms per mL (in the units often used). The measured values we have are all lower than this save for MMV670944. (note solubility values on wiki wrongly in mg per mL - will change)

[MFernflower]

@mattodd @MedChemProf

[edwintse]

With regards to the LH/RH pyridazine compounds suggested by Chris Swain above, the LH pyridazineethanol is available commercially however it seems that it is quite pricy and the lead times are quite long (https://chem-space.com/search/58251f35-5a8f34-814e38b1-08e3df1/CSC010984883.html?currency=usd http://www.accelpharmtech.com/search.do?a=s&searchtype=2&psize=50&q=101479-62-3&searchtmp=goodssearch).

The RH pyridazinealdehyde can be purchased from Matrix Scientific as either the 4-Cl (http://www.matrixscientific.com/102580.html) or the 4-CF3 (http://www.matrixscientific.com/090030.html) derivatives

[mattodd]

Awesome, Ed. What say you @drc007 ? The RH compounds seem a no-brainer. Any alternatives to the LH?

[drc007]

For the right-hand side the available pyridines seem good choices. Since key question we are trying to address is "Can we increase solubility whilst maintaining potency" for the left hand side I'd just use a simple phenethyl ether since this would allow direct comparison with the corresponding 4-chlorophenyl and 4-CF3phenyl analogues.

[drc007]

For the left-hand side 3-Bromopyridazine is commercially available from many vendors (but I could not find cost), perhaps use this and build in linking group in the first instance?

[edwintse]

I have 2 possible ways to access the LH pyridazine alcohol.

The paper describing the first reaction scheme below (Afr. J. Pure Appl. Chem. 2011, 5(6), 123-126.) reports the synthesis of 2-phenylethanol but should be applicable to pyridazine. 3-Bromopyridazine can be purchased from Matrix Scientific as well. The only issue is with the use of 1,2-dibromoethane, which is a restricted carcinogen.

The second route is shown below is from two different papers (10.1248/cpb.27.916 and 10.1055/s-1990-21045 for the first and second steps respectively). 3-Methylpyridazine can be purchased from Sigma or Matrix.

[MFernflower]

@edwintse What about ethyl chloroacetate as a replacement for the 1,2-dibromoethane???

[edwintse]

@MFernflower Yes those could work, then reduce after.

[MFernflower]

@edwintse @drc007 @mattodd Current proposed synthetic route for LH alcohol:

[mattodd]

Let's order the aldehyde in the meantime, for the RHS (action on @edwintse ). Ed I also have 10.1039/c39850001632 as a possible lit source for the synthesis of the alcohol, according to Scifinder?