When has unsupervised clustering found anything interesting?

At Teiko, we used unsupervised clustering to identify distinct Treg functional states associated with better survival in 40 patients with urothelial carcinoma.

We initially applied both gated and unsupervised clustering analysis to identify immune features associated with response. Gated analysis found that higher frequencies of regulatory T cells (Treg) were associated with better survival, a surprising result given that Treg usually suppress tumor-killing immune cells. Unsupervised clustering split the Treg population within the dataset into two distinct clusters. Only one of those was associated with response, and that cluster showed Treg were inactivated. More inactivated Tregs potentially results in stronger anti-tumor activity and greater response to immunotherapy for patients. This crucial Treg population would have been really difficult, if not impossible, to find with manual gating alone. Click the button below to see how we identified this unique Treg cluster!