Nearly 25 years after the first proteome arrays were conceived, the platform that began as a way to assess antibody specificity has expanded into a powerful discovery engine. In this second article, Ignacio Pino reflects on how CDI Labs moved beyond validation into autoantibody seromics, the body's natural antibody repertoire and interactomics, and what that evolution may mean for the next generation of drug discovery.

Discovery at Scale: High-Content Biology and the Next 25 Years of CDI Labs 

In our previous article, we described how CDI Labs evolved from an antibody specificity validation platform into an autoantibody seromics discovery engine. But what does that mean in practice? 

Proteome-scale screening is often misunderstood. One of the biggest surprises over the past two decades has been how frequently researchers expect a high-content platform to deliver a final clinical answer. But it doesn’t - it was never designed to. 

A proteome array is a discovery engine. You start wide, identify candidate signals, validate those signals in larger cohorts. Then you refine them into focused assays suitable for clinical or translational use. The array is the first step in a journey, not the final destination. 

Understanding that progression has been one of our ongoing educational challenges. 

The “Jewel” of the Platform 

At the core of CDI Labs is a functional human protein collection. 

Our collection contains more than 21,000 full-length proteins, individually expressed and subjected to strict quality control.  Each protein is expressed in full-length, correctly folded and has undergone post-translational modification. The result is the retention of enzymatic activity and conformational integrity. 

Because the collection is replenishable and rigorously validated, it can be deployed in multiple contexts. The slide format is one application. In other cases, transferring the proteins into liquid-based systems provides better performance. The underlying asset is the collection itself. 

That collection enables multiple pillars: antibody specificity validation, autoantibody seromics, mining the natural antibody repertoire, and increasingly, exploration of protein–protein interactions. 

The Natural Antibody Repertoire: Learning from Exceptions 

One of the most promising areas we have explore is the antibody profiles of samples from within or across cohorts. 

Instead of designing antibodies entirely from scratch, we examine samples where response has been exceptional to a therapy. Are some immune systems already generating highly specific, functionally effective antibodies? In the research environment, using proteome-scale screening, we can identify those responses, isolate B cells, and sequence the corresponding antibodies. 

In a sense, we are working with solutions that evolution has already refined. 

This approach reduces guesswork. Rather than relying solely on affinity as a selection criterion, we assess specificity across the entire proteome. An antibody can bind tightly to its intended target and still cross-react elsewhere if similar epitopes exist. 

Proteome-wide screening allows us to see that broader landscape of antibody specificity. 

From Proteomics to Interactomics 

As the field has matured, our perspective has expanded from individual proteins to networks. 

With more than 21,000 canonical proteins, the potential interaction space within the human proteome is vast. Understanding how proteins interact, and how those interactions fail, opens new therapeutic avenues. 

We are now entering what many describe as the era of drugging the undruggable; rewiring the protein:protein interactions, where molecular glues are a very attractive alternative. 

One of the most important next steps in drug discovery is finding ternary complexes that allow the ‘undruggable’ proteome to be targeted – that's approximately 90% of the proteome. Compared with biologics, small molecules are generally less costly to manufacture, which has implications for accessibility and scalability in healthcare. 

Mapping interaction networks with precision will be central to this effort. A functional proteome collection provides a foundation for that work. 

Data, Specificity, and the Role of AI 

Artificial intelligence is reshaping biomedical research, but its effectiveness depends entirely on the quality of data it is trained on. 

Across our antibody specificity validation and autoantibody seromic discovery campaigns, we generate thousands of data points per experiment. When antibody sequence information is paired with full proteome-scale specificity profiles, it becomes possible to train systems to distinguish truly selective antibodies from those that merely exhibit high affinity. 

Much of this data already exists within our archives. The next step is systematic curation and integration. 

In that sense, CDI Labs can train and validate AI systems, particularly in the domain of biological specificity, where incomplete datasets have historically limited predictive accuracy. 

That work builds on the same principle that shaped the company from the beginning: start with a comprehensive, functional view of the proteome and let the data guide the next step. 

The Next 25 Years 

CDI Labs began with an ambitious objective: represent the human proteome in a functional, interrogatable format. 

The platform didn’t change all at once. We first applied it to proteome-scale antibody specificity because that was the problem we could solve. As new datasets emerged, we moved into autoantibody seromics and later into molecular glues. Each step followed practical evidence rather than a fixed roadmap. 

The tools will change further over the next 25 years. Computational methods will improve and therapeutic strategies will evolve. What is less likely to change is how we approach the science. We start broad, examine the full proteome, and narrow only after the data justifies it. 

That consistency has shaped CDI Labs from the beginning.