One selection campaign delivering both hits and biology is a nice outcome. Macrocycles sitting on a protein–protein interface are rarely drug-like as they stand, so my question is whether the ternary structure lets us shrink them toward something cell-permeable; the second site on CHD4 gives another handle for that. I'd also want to see selectivity against related remodelers.

Duration is the headline, but the permanently charged ammonium is the more interesting design choice: it should slow washout from the injection site, which probably explains the long action. For anything that lasts over 50 hours, though, I'd look at local neurotoxicity and myotoxicity data before the efficacy curves.

Locking a floppy piperidine into a fused bicycle is the textbook way to sharpen selectivity, and it seems to have paid off here. Before getting attached I'd check what the extra carbon framework costs in lipophilicity and brain exposure, because a CNS target punishes both. Mutant-over-wild-type selectivity also needs to hold up in cells, not just in biochemical assays.

STAT6 is a transcription factor with a flat, polar SH2 surface, exactly the kind of target where inhibitors struggle and degraders can sidestep the problem. Getting oral exposure from a PROTAC that started life as a phosphopeptide is the real achievement; I'd like to see how molecular weight and polarity were managed, and whether the STAT-family selectivity holds in primary T cells.

Using the γ-phosphate to position the warhead is neat: GTP-state selectivity comes from the nucleotide itself, not only the pocket. The comparison I'd want is against RAS(ON) tri-complex inhibitors, which need cyclophilin A while this doesn't. Acrylamide reactivity and proteome-wide cysteine selectivity also matter more here than for GDP-state binders.

Lysine-directed covalent chemistry is far rarer than cysteine, so a workflow that actually delivers cell-active hits on a lysine is the takeaway, more than SHP2 itself. My caveat is reactivity versus stability: lysine electrophiles can be hydrolytically fragile or reversible, so I'd want proteome-wide selectivity data before calling anything a probe.

DEL-trained models that shine on random splits and collapse on new chemotypes is exactly what bites when you try to seed a different series from a DEL campaign. Three targets is a small panel, but a like-for-like comparison with physics-based methods on identical data is useful. I'd use it to calibrate how much to trust ML triage on scaffolds the library never sampled.