Natural products and the first turning point
In its early stage, the Institute drew inspiration from natural-product medicine. We used image-based phenotypic screening to evaluate medicinal herbal extracts and their fractions, isolating natural compounds with selective activity against defined oncogene-driven cellular states, with particular focus on MYC-driven, high-demand malignant regimes. This work produced multiple validated leads and clarified an essential point: the activity landscape of natural products is not “finished” simply because many compounds have been studied before. New assay systems can reveal biological functions that were not previously tested, even in well-characterized molecules. What appears “known” often reflects historical assay coverage rather than true exhaustiveness. The limiting factor is often not the chemistry itself, but the sensitivity and framing of the assays used to interrogate it.
A translational constraint and the move to tractable chemistry
As we moved toward translation, we confronted a practical constraint: the structural complexity of many natural products can make synthesis and systematic optimization for drug-like properties difficult. Phenotype-first discovery introduces a second constraint: when a cellular phenotype is compelling but the molecular target is not yet defined, identifying synthetic compounds that reliably reproduce the natural-product phenotype is a hard search problem. To preserve phenotypic sharpness while enabling tractable medicinal chemistry under these conditions, we developed GUNS-DF (an on-demand, selective screening library) to deliver actionable starting points with minimal experimental burden. The approach supports identification of phenotype-mimicking hits from a small set of synthesized compounds, followed by rapid optimization to high potency within a few medicinal chemistry cycles, typically requiring hundreds of additional synthesized analogs rather than large-scale campaigns.
Convergence and the necessity of an architectural explanation
Across this body of work, we repeatedly observed a decisive pattern: structurally unrelated scaffolds converged on the same state-level disruptions, including reproducible organelle fragmentation, loss of trafficking coherence, and centrosome fragmentation and declustering. This convergence was not parsimoniously explained by isolated target models, because distinct chemistries associated with different nominal targets were collapsing the same cellular state. Our working resolution was a deeper governing layer: malignant systems remain viable only while their internal organization remains executable under load. When organization is pushed beyond a constraint boundary, collapse can occur even when classical signaling interpretations remain incomplete. This inference became the foundation for the Institute’s conception of Architectural Oncology: a framework for identifying and testing architectural vulnerability in malignant states, without assuming that all drugs or diseases operate by the same logic. Where target-dominant explanations are sufficient, we use them; Architectural Oncology is invoked when state-level convergence indicates a governing constraint.
From theory to testability: SYMPLEX
A theory of architectural vulnerability requires disciplined stress testing. The Institute therefore developed SYMPLEX (an integrated discovery-and-stress-test system), a recursive, ascending triplex in which each cycle raises the constraint standard and advances only what survives progressively stricter tests, refining both the compound series and the architectural hypothesis under evaluation. In SYMPLEX, GUNS-DF provides controllable chemical diversity; MIPS (mechanism-informed phenotypic screening) quantifies reproducible organelle- and state-level disruption; and DrugGPT supports disciplined interpretation and documentation of hypotheses so that claims remain within predefined admissibility and falsifiability criteria. SYMPLEX is designed to earn architectural claims through predefined, progressively stricter stress tests, not to decorate hits with narratives.
Structural Synthetic Lethality and translational relevance
This evolution refined our therapeutic strategy into Structural Synthetic Lethality (SSL): an approach that extends beyond traditional genetic by targeting architectural vulnerability, defined as collapse of a viable malignant configuration rather than single-gene dependencies.
Today, the Institute advances basic research and theory-building alongside translationally relevant validation. Within the same Institute, we developed the framework, built the stress-test system that makes it falsifiable, and advanced multiple scaffold-distinct small-molecule series as rigorous probes of architectural vulnerability in MYC-driven cancers.
