“Or it’s real, and it’s been used. Eight hundred ninety-two subjects. That’s not a lab study, Maya. That’s a clinical trial. A very illegal, very clandestine one. And v1.9.10 means there were nine iterations before this. Nine chances to kill people.”

But the version number—v1.9.10—suggested it had been refined. Iterated. Tested.

The 0.4% all had the same rare HLA variant—HLA-B 57:03, a known anomaly. The notes table had a partial entry for one of them: “B 57:03 escape variant. v1.10 in progress.”

They agreed to run a virtual validation. Kettering had anonymized HLA data from 10,000 transplant patients. Maya wrote a script to simulate the “Fresh Supply” protocol on a subset—just in silico, just predicting rejection probabilities.

No. Not just transfusion. Transplantation. Whole organs, tissue grafts, bone marrow—without matching. Without the lifelong cocktail of anti-rejection drugs that left patients vulnerable to infection, cancer, kidney failure.

Someone had leaked this. Someone on the inside.

If this was real, it was the Holy Grail of transplant medicine.

It was a file name like any other on a Tuesday afternoon—until it wasn’t.

They were still iterating. Maya dug deeper into the supplemental.bin file. It wasn’t binary in the usual sense—it was a compressed image. When she extracted it, she found a single photograph: a hand-labeled freezer rack. On each cryovial, handwritten in black marker:

Maya felt a chill that had nothing to do with the lab’s HVAC. She opened main.db .

She opened schema_v1.9.pdf . Forty-seven pages of dense immunogenetics, but the summary diagram stopped her cold.

She looked down at her arm, at the small white scar from the donation needle.

Predicted rejection rate without protocol: 68% (for mismatched donors). Predicted rejection rate with protocol (v1.9.10): 0.4%.

And anyone could have taken her HLA profile.

She should have flagged it for the encryption alone. Open science was the rule in pathogen genomics. Unbreakable encryption meant someone had something to hide. But the system didn’t auto-flag because the header wasn’t malicious—it was just… strange.