Drug Discovery Shifts to New Targets: RNA, Cell Surface Proteins, and Sequence-Based AI

The scientific community has made tremendous progress in the ability to identify and predict protein targets, advancing from bulk cellular analysis approaches to technologies with resolutions that now approach subcellular. Despite these achievements, fundamental questions in drug discovery remain difficult to answer: has the ideal protein target been identified, and is it easily accessible to intervention to accelerate the development pipeline and reduce candidate attrition?

Today, it's estimated that 60% to 70% of all drugs on the market target cell-surface proteins, including monoclonal antibodies, antibody-drug conjugates, and CAR T-cell therapies. This is despite the fact that the underlying drug discovery work for those therapies was done without a tool designed to focus specifically on cell-surface proteins.

Traditional protein targets are becoming increasingly saturated, driving interest in alternative approaches. RNA-based therapeutics have garnered much attention in recent years. An antisense oligonucleotide therapy was approved for spinal muscular atrophy in 2016 and the first RNA interference therapeutic gained approval in 2018. In addition, the 2024 Nobel Prize in Physiology or Medicine was awarded to the discovery of microRNA and its role in post-transcriptional gene regulation.

However, RNA-based therapeutics have often been hindered by delivery challenges due to the large, hydrophilic nature of RNA and its susceptibility to degradation. It is argued that RNA-targeted small molecules could offer the same transcription-level intervention, while providing the added benefits of oral availability and scalable manufacturing.

Traditional small molecule drug discovery is predicated on the identification of well-defined binding pockets, which is not conducive to the dynamic nature and relative thermodynamic instability of RNA in comparison with proteins. Nonetheless, advances in the understanding of RNA structural biology and high-throughput screening techniques have allowed for the identification of RNA–small molecule binding interactions. The key challenge has now evolved from identification of RNA binders to enhancing RNA selectivity.

A succession of partnerships with biotechs focused on RNA-modulating small molecules have signalled big pharma's growing interest in the space. In 2025 alone, Merck KGaA announced a collaboration with Skyhawk Therapeutics in a deal worth up to $2bn; Daiichi Sankyo partnered with Wayfinder Biosciences for use of its drug discovery platform in neurodegenerative disease and Astellas Pharma revealed plans to collaborate with xFOREST to utilise its RNA splicing-targeted drug discovery platform.

Sector progress has been driven, in part, by the landmark success of an oral SMA drug which first received FDA approval in 2020. SMA is characterized by deficient SMN protein. The drug binds two sites on exon 7 of the SMN2 pre-mRNA—namely, ESE2 and 5'ss—to promote their inclusion in the mature transcript, thereby increasing functional SMN protein levels.

Clinical-stage Remix Therapeutics collaborated with Johnson & Johnson for exclusive rights to three specific targets in immunology and oncology, for a $45m upfront payment and other payments potentially exceeding $1bn. In January 2024, Remix formed a partnership with another company for the discovery and development of small molecule therapeutics modulating RNA processing. The deal included a $30m upfront payment and up to $1.12bn in milestone payments and royalties.

In the cell surface protein space, these proteins serve as the primary communication and regulatory interface between a cell and its environment; they act as the actionable biology gateway for therapeutic intervention. Cell surface proteins, which collectively can be thought of as the "surfaceome," are notoriously difficult to study for a variety of reasons. Most proteins aren't static on the cell membrane. They are transient, actively changing in response to cell state, environmental stimuli, and disease.

While plasma membrane proteins comprise about 2% of total protein abundance, they are highly important and actionable, as showcased by the number of FDA-approved drugs targeting cell membrane proteins. These proteins are hydrophobic, heterogeneous, and often low in abundance, making them difficult to isolate. Since the functional cell surface is only 10 nm to 50 nm, analysis tools must have exceptional spatial precision and sensitivity to accurately capture surface proteins without all the noise caused by contamination from intracellular proteins.

In target identification, success is defined by both biological relevance and physical accessibility. Surfaceome analysis focuses on disease-associated proteins positioned at the cell membrane where therapeutic agents can directly engage their targets, maximizing the likelihood of efficacy. For biomarker discovery, surfaceome-associated proteins are ideal biomarkers because they are often biologically relevant, clinically accessible, and provide more than just measures of disease.

Ainnocence announced a major milestone in artificial-intelligence driven drug discovery: a sequence-first AI platform capable of screening billions of small-molecule and antibody candidates across the entire genome in hours on a single GPU without relying on 3D structural modeling. This approach represents a fundamental shift in how therapeutic candidates are discovered, replacing decades of structure-dependent simulations and computationally expensive systems with direct learning from biological sequences and experimental data.

Traditional AI drug discovery pipelines are constrained by protein structure availability and limited 3D structure prediction, computationally intensive molecular dynamics, and limited throughput. By contrast, the platform operates entirely at the sequence level, enabling whole-proteome virtual screening without solved structures, billions of candidates evaluated in hours, not months, 80% reduction in wet-lab cost and time, and 10–60% experimental hit rates, far exceeding industry averages.

The platform has been applied across 60+ therapeutic programs, including antibodies, small molecules, cell therapies, siRNA, and synthetic biology applications. In one landmark study, Ainnocence computationally designed mutation-resistant antibodies against SARS-CoV-2 variants, successfully predicting neutralizers for Delta and Omicron before Omicron emerged, demonstrating the model's ability to learn evolutionary patterns directly from sequence data.

Internal benchmarks show that protein foundation models achieve Spearman correlation performance comparable to structure-based models, while requiring orders of magnitude less compute and running efficiently on a single GPU. This breakthrough has been featured in Chemical & Engineering News, the flagship news outlet of the American Chemical Society.