SCHEMBL3899520

Similar compounds — the chemically nearest patent molecules

Nearest neighbours by Morgan-fingerprint cosine across the patent-compound collection, with each neighbour's top predicted target and the predicted targets it shares with this molecule.

	Compound	similarity	top predicted	shared targets
	SCHEMBL3903838	0.90	BTK (0.46)	DPP4BTKJAK3RETMAP4K4
	SCHEMBL4564489	0.90	BTK (0.46)	DPP4BTKJAK3RETMAP4K4
	SCHEMBL920704	0.90	DPP4 (0.44)	DPP4BTKJAK3RETKDM4D
	SCHEMBL5434992	0.84	DPP4 (0.52)	DPP4BTKJAK3KDM4DMAP4K4
	SCHEMBL919746	0.83	DPP4 (0.43)	DPP4BTKJAK3RETKDM4D
	SCHEMBL12906394	0.83	DPP4 (0.45)	DPP4BTKJAK3RETKDM4D
	SCHEMBL5433194	0.81	DPP4 (0.52)	DPP4BTKJAK3RETKDM4D
	SCHEMBL13647417	0.80	DPP4 (0.53)	DPP4BTKJAK3RET
	SCHEMBL13647468	0.79	DPP4 (0.52)	DPP4BTKJAK3RET
	SCHEMBL16313495	0.78	DPP4 (0.53)	DPP4BTKJAK3RET

Similarity is cosine over the 2,048-bit Morgan fingerprint (≈ Tanimoto). Identical fingerprints score 1.00.

Patent provenance — the patents this molecule appears in, and who filed them

Claimed or disclosed in 10 patents. claimed = in the patent's claims; disclosed = body only.

Patent text — is the patent's own abstract consistent with the prediction?

For each of this compound's patents that has machine-readable text (3 of them — usually the abstract, not the full specification), we ask MedCPT which protein the text reads most about, and where the chemistry-predicted target lands among 4885 human targets. A high rank means the patent's own wording is consistent with the prediction — a weak, independent signal, not proof of activity.

“Text reads most about” is the patent abstract's nearest protein in MedCPT space (background-debiased). Only ~1.4% of patents have machine-readable text, so most compounds won't have this panel.