A couple of safety issues

I’ll flag up a couple of safety issues mentioning @frankvondelft @MarkC @JohnChodera @edgriffen @alphalee @londonir @Tanya @Ben_DNDi @mc-robinson

Currently the design team are looking for P1 substituents with improved metabolic relative to isoquinoline. The aromatic amines from which the P1 substituents are derived would raise generic concerns (amino group attached to aromatic carbon) about genotoxicity. Pharma companies typically have in house genotoxicity data (e.g. Ames assay results) as well as models trained using their proprietary data. While I don’t think that Pharma companies will make all their data publicly available, I’m guessing that they would be prepared to check databases and run models for specific aromatic amines. My understanding is that there are a number of people working in Pharma who are associated with the COVID Moonshot and I would recommend that you explore possibilities with them.

The majority of the inhibitors of interest in the 3-aminopyridine series are chiral and this is a potential safety/development issue if the configuration is not locked (e.g. with methoxy) since the chiral center is next to the amide carbonyl. As discussed previously, it is possible that locking the configuration of the chiral center could alter the existing SAR. This is most likely to be an issue for tetrahydroisoquinolines at P2 with relatively large substituents on nitrogen such as MAT-POS-dc2604c4-1 | EDJ-MED-1981ceba-4 | MAT-POS-4223bc15-23. I would recommend that you assess the various configurational locks of interest sooner rather than later if you think that it’ll be necessary to lock the configuration of the chiral center.

@pwkenny thanks for the notes. Mini-AMES is on the cascade critical path (and full genotox for later candidate(s)), and we could have a couple of compounds pushing towards that stage sometime soon. My experience with the “aniline” question is that it’s considered less of an issue for direct aromatic amines where the aminosubstituted aryl group is heteroarmatic (as in the case of the isoquinoline), although that viewpoint may have changed (been a while). We have a couple of pharma companies working with us who may be able to scan their internal “acceptable anilines” data bases should they exist. As always with these things it’s the experimental which is needed to validate the theory.

For the racemization risk, Ed may be better placed to comment but my understanding is that that centre is a little sensitive in basic media, but not too much of an issue at neutral or acidic pH. Again it’s about testing stability in the relevant media for the most interetsing compounds. There is a plan to try to use quartenary centres to avoid the issue at all, but that brings synthetic compexity which we need to manage.

Hi @Ben_DNDi

With respect to potential for genotoxicity, I think that it’d be good to get as many eyes as possible looking at the aromatic amines. With luck, you may even find people are prepared do assessment that goes beyond simply looking in databases and running models. The challenge is how best to connect with those people who may have heard of the COVID Moonshot but are thinking that they have nothing to offer. While I’d anticipate that aza-substitution of an aromatic amine would reduce potential for genotoxicity, I think that the aza nitrogen of 4-aminoisoquinoline is not in the best position for doing this.

Even if the unlocked configuration doesn’t cause specific problems (e.g. inactive enantiomer is toxic), it could complicate development if the regulators deem it necessary to show that racemization doesn’t happen in vivo and/or that the inactive enantiomer is clean. My understanding (I’ve not checked) is that thalidomide racemizes in vivo although its chiral hydrogen is likely to be more acidic than those in the COVID Moonshot compounds. In any case the methoxy configurational lock also appears to confer an advantage in the antiviral assays (I don’t know if this is still the case). These notes on conformational preferences of alkoxy on carbon next to amide carbonyl may be of interest.

Hello all! I’ve been taking a look at the Ames issue using the QM protocol we published in 2012 (DOI: 10.1021/jm3001295). So far, I have looked at a few substituents at the 7-position and computed probability of being active:
H: 43.2%
Cl: 36.5%
F: 37.4%
NHSO2Me: 43.8%
SO2Me: 32.5%
I’m going to try and do the same for a few more groups and at the 6-position in the next couple of days.

I’ve done some further calculations on other variants.

For 7-substituents
CF3: 33.7%
CH2N=S(=O)Me2: 56.9%
N=S(=O)Me2: 60.5%
CONHMe: 41.4
CONHcPr: 40.0

For 6-substituents
H: 43.2%
Cl: 41.1%
F: 42.2%
SO2Me: 32.2%
CF3: 33.3%

Alternative heterocycles
Nc1cncc2c1[nH]nc2: 39.5% (two tautomers considered)
Nc1cncc2c1[nH]cc2: 48.6%

Hi @anandrewleach these probabilities all seem pretty high. Should this be taken as an indication that all should be assayed for genotoxicity?

I think you are right that particularly the ones at the upper end would require some real world checking if you were considering taking them forwards.

I’ve been exploring some other variations and hope that my nomenclature makes sense.

5-Cl: 42.1%
5-Cl, 7-SO2Me: 32.4 %
6-O-alkyl: 60.8 %
6-C(Me)2OH: 50.7 %
6-CONHMe: 42.4 %

Other heterocycles:
5-Cl, 6-aza: 27.8 %