Since the development of DNA microarrays over thirty years ago, and accelerating with the advent of RNA sequencing, it has been possible to measure the expression of almost every gene in a biological sample. This led to the identification of genes whose expression was altered in a diseased tissue versus a healthy tissue, and these differentially regulated genes and signaling pathways often made for good drug targets. The era of pathway-targeting pharmacology was born.

However, this information came at the bulk sample, not the individual cell, level. In contrast to the relative ease of measuring gene expression, it was nearly impossible to thoroughly catalog the different cell types and states present in a tissue, let alone their differential abundance in diseased versus healthy samples. Technologies like flow cytometry and immunohistochemistry can identify different types of cells, but only in a hypothesis-driven fashion. This requires scientists to pre-select a small set of detection antibodies and makes cell type identification a low-resolution process. Consequently, it was intractable to identify and target specific cell populations in most diseases.

The development of single cell sequencing technologies has changed that. It is now possible to measure the gene expression profile (and increasingly, the protein expression profile) of every individual cell in a sample, without any prior hypothesis about which genes are expressed. Cells can then be computationally grouped into cell types and states based on their similarity. The picture that arises from this technology is like putting on corrective lenses: blurred lines become sharp, and details that were invisible become trivially easy to see. Historically monolithic cell populations such as fibroblasts and epithelial cells separate into different cellular subsets: each is a type of fibroblast or epithelial cell, but the subsets have distinct profiles and functions. In this context, it becomes possible to identify specific pathogenic subpopulations; whereas eliminating all epithelial cells would be toxic, eliminating a specific subpopulation of dysfunctional epithelial cells may ameliorate disease without safety concerns.

At Arda, we use custom machine learning methods integrate single cell data across studies, technologies, and between patients and precinical models. Our target selection platform maps disease-specific cell states, then incorporate selectivity, robustness, and translatability metrics to select surface targets. We then validate targets experimentally, develop biologics, evaluate efficacy and safety, and move forward to drug development.

Once a pathogenic cell population is identified, the next step is to develop a therapeutic that eliminates those cells, and only those cells. This is more easily said than done. Even in oncology, in which the target cells are quite distinct from their normal counterparts, many therapies come with substantial collateral damage. Most cancer drugs take advantage of metabolic differences between cancer cells and normal cells; for example, preferentially killing rapidly dividing cells by exploiting the biology of cell replication. This strategy of targeting intracellular vulnerabilities comes with two key limitations: 1) one must understand the internal wiring of the cell well enough to identify the weak points, and 2) those mechanisms are rarely unique to the target population; for example, some normal cells are highly proliferative, hence the dramatic side effects of chemotherapy.

In the last ~10 years, however, the oncology therapeutic toolbox has added new modalities. Immune-cell engaging biologics, antibody-drug conjugates, and engineered T cells all localize cytotoxicity to specific cells based on the expression of one or more cell surface factors. A deep understanding of the internal vulnerabilities of the target cell is not needed; these highly potent agents require only a homing beacon.

These approaches have shown dramatic clinical efficacy, and we believe their potential extends beyond oncology. Clinical and preclinical data has shown us that these modalities can eliminate non-malignant cells in a target-specific manner. With the right choice of surface factor, we can target virtually any pathogenic cell population for elimination. The tools of immuno-oncology may actually work better outside of oncology, as cancer poses a number of unique challenges that are avoided in other diseases, such as evolved resistance and the need to eliminate virtually 100% of target cells.