AML drug development is running two completely different playbooks, and one of them is still unproven.

We used Sleuth to map 953 active AML programs. When you split the landscape by how targets are drugged, the divergence is striking.

Kinase and enzymatic targets like IDH1, Bcl-2, and menin have produced most of the modern targeted AML approvals, from venetoclax to revumenib to ziftomenib. These 47 programs across 5 targets are drugged almost entirely with small molecules and selected by clear biomarkers. FLT3 reinforces the pattern: a surface receptor drugged as a kinase, 89 programs, 3 approved.

Surface antigen targets like CD33, CD123, CLL-1, and CD7 tell a different story. These proteins sit on the outside of the leukemic blast, accessible to cell therapies, ADCs, and antibody-based approaches. Across these four targets, 118 of 162 target-program pairs are cell therapies, most of them CAR-T. Of 121 unique surface-targeted cell therapy programs, exactly one has reached Phase II. Twenty more sit in Phase I/II, but none have generated pivotal data. This pipeline is large and early, not failed, but overwhelmingly unvalidated.

The modalities with actual AML validation against surface targets are barely represented. The only approved surface-targeted AML drug is gemtuzumab ozogamicin, an ADC against CD33. There are 9 surface ADC programs total, 3 in Phase II or beyond. A handful of antibody-based programs have also reached Phase II, though several are drugs approved in other indications being tested in AML. Cell therapies outnumber ADCs 13 to 1 despite zero AML approvals.

Some of this reflects timing. Kinase targets have had decades of development while surface antigen cell therapy is largely post-2018. But if the field recognized these targets are early, you would expect diversified modality bets. Instead, 61% of surface-targeted programs are cell therapy, concentrating risk in a single unvalidated approach while the modality with the only approval in the class gets a fraction of the investment.