Finally, we teamed up with @pschwllr.bsky.social @rebeccaneeser.bsky.social who prepared a random forest model that can readily applied to estimate the membrane permeability of designed cyclic peptides.
Model available on GitHub: github.com/schwallergro...
(8/🧵)

This gave us deep insight into the permeability of an unprecedented number of compounds in CAPA.
We also recapitulated that exceeding a MW of 800 Da, a PSA of 250 Å2, or 5 HBDs had a small chance of having a good membrane permeability - although there can be outliers. (7/🧵)

After some synthesis iterations, we designed a diverse random library of 192 CA-peptides. Despite some issues with diketopiperazine (DKP) by-products, we obtained 143 macrocycles for subsequent crude analysis in CAPA. (6/🧵)

We next developed a crude synthesize strategy to synthesize crude CA-tagged macrocycles (based on previous workflow: onlinelibrary.wiley.com/doi/10.1002/...), and we saw that crude peptides tested in CAPA gave rise to similar half-maximal cell penetration (CP50) values to purified compounds. (5/🧵)

To make CAPA directly compatible with automated SPPS, I synthesized new (and less polar) CA-tagged amino acids that allowed for easy and direct incorporation into macrocycles. We showed that permeability was nicely retained in known model peptides such as CsA bearing the new amino acid(s). (4/🧵)

One of the caveats with CAPA for passively permeable compounds is the relatively large (commercial) succinyl-chloroalkane tag (Suc-CA).
Permeability was negatively impacted >10-fold compared to installing a smaller CA-tag onto macrocycle scaffolds with known good passive permeability. (3/🧵)