Contact

KiraLOGIC

Multiplex editing expanded what can be built. Choosing which edits belong together is the next bottleneck.

KiraLOGIC is our experimental design engine for multiplex-edited cell therapies. It combines computational predictions with results from cells we have built and tested to prioritize which combinations of gene edits to evaluate next.

Why KiraLOGIC

The product is the combination.

Within a defined therapeutic backbone, the edits have to work as one coordinated architecture.

Genome-wide screens, spatial and clinical datasets, and mechanistic studies can all point to promising targets. But knowing the right targets does not tell you which ones belong together.

Once edits are combined in the same engineered cell, their effects do not simply add. They can interfere with one another, weaken the cell, or become more powerful together.

The tumor does not test those edits one at a time. It pushes back on the engineered cell as a whole.

From evidence-weighted targets to a focused experimental setMore than 80 evidence-weighted targets form more than 30 billion theoretical four- to eight-target combinations. Program constraints and KiraLOGIC narrow that space to a focused experimental set matched to assay capacity.

Before anything is built, we constrain the search by the program’s biology and goals, build feasibility, and assay capacity. KiraLOGIC then proposes a focused experimental set.

KiraLOGIC does not replace the experiment. It helps make each experiment count.

How KiraLOGIC works

Start broad. Learn from every build. Choose what to test next.

KiraLOGIC combines data from KiraGen’s engineered cells with Phenomena, Reflector Bio’s pretrained model of cellular response.

Each experimental record connects the intended combination of edits, the cells actually produced, and their measured function under defined conditions.

That evidence guides the next experiments: testing promising combinations, resolving important uncertainties, and exploring different designs within the program’s biological goals and available laboratory capacity.

About the underlying model

Current Phenomena corpus figure supplied by Reflector Bio: 21 million experimental conditions. “Experimental conditions” is Reflector's reporting unit. Corpus scale establishes provenance, not KiraLOGIC performance.

The KiraLOGIC experimental learning loopA broad functional prior and measured-build reference feed KiraLOGIC. Each record links the intended architecture, realized build, and measured function. KiraLOGIC prioritizes a proposed experimental set, builds are tested, and qualified results update the reference. Blue paths show information flow and green paths show evidence flow.
Propose → Build → Test → Learn

How we test

Test broadly. Use KiraLOGIC to decide where to go deep.

Different experimental models answer different biological questions. We compare a broader set of architectures in scalable assays, then move only a few into models that provide the context needed for the next decision.

Each qualified result informs what KiraLOGIC recommends testing next and where. This makes resource-intensive models part of selection, not just final validation.

KiraLOGICFocuses scarce experimental capacity on the next decision
Experimental capacity allocated across complementary assay contextsA broad, nonliteral field of candidate architectures is allocated across controlled in vitro assays, patient-derived spheroids, patient-derived tumoroids, and orthotopic PDX models as the program decision narrows. The diagram does not represent fixed build counts or a mandatory serial ladder.

KGEN-001 reference

KGEN-001 anchors our first GBM reference.

KGEN-001 was defined using human glioblastoma biology before KiraLOGIC was developed. Data from precursor and related cell designs now form the initial GBM reference used to prioritize future combinations.

Explore KGEN-001
Measured cell designs feed a GBM reference that informs future combinations of edits.
Engineered T cells surrounding a tumor spheroid in a defined assay chamber.

Collaborate with us

Bring us a combination problem.

Bring a cell-therapy program and a defined biological challenge.

Together, we can prioritize combinations of gene edits and test how they affect cell function in relevant models.

Discuss a KiraLOGIC collaboration