Allogeneic iPSC-derived islets normally trigger rejection because donor HLA molecules are seen as foreign. A "hypoimmune" graft edits that recognition machinery instead of suppressing the whole immune system (as VX-880-style trials still require today):
P_Tcell = baseT · (HLA-I? 1 : 0.05) · (HLA-II? 1 : 0.55)
· (1 − 0.75·PD-L1) · (1 − suppression)
P_NK = baseNK · (HLA-I? 0.12 : 1) · (1 − 0.8·CD47)
· (1 − 0.3·suppression)
P_kill(day) = 1 − (1 − P_Tcell)(1 − P_NK)
Knocking out HLA-I removes the ligand CD8⁺ T cells scan for — but it also removes the inhibitory signal NK cells rely on ("missing-self"), so NK attack rate rises unless CD47 is added to trigger the SIRPα "don't eat me" checkpoint on NK cells and macrophages. HLA-II knockout starves CD4⁺ help; PD-L1 exhausts whatever T-cell response gets through. Real hypoimmune platforms (e.g. B2M/CIITA knockout + CD47 + minimal HLA-E) combine all of these rather than relying on any one edit.
Graft function is not binary — it ramps up as the vascularising islet matures, then falls if immune attrition outpaces it:
maturity(t) = min(1, t / 14 days)
insulin output = (alive fraction) × maturity(t)
- Toggle the four edits — each multiplies into the daily kill probability above; watch T cells (red) and NK cells (violet) probe the islet cluster and see which combination stops them.
- Immunosuppression slider — the conventional alternative: dampens both arms broadly instead of editing the graft, at the cost of systemic drug exposure this model doesn't otherwise charge for.
- Re-graft — reseeds a fresh 36-cell islet cluster and restarts the clock under whatever protocol is currently set.
Coefficients (baseT ≈ 4%/day, baseNK ≈ 2.2%/day) are illustrative teaching parameters tuned to make the qualitative trade-off visible in minutes, not measured clinical rejection rates.