Bridging Probability of Success and Early Drug Discovery in Animal Health

In the world of animal-health drug discovery, success is elusive and often capricious. The vast majority of experimental products will never reach the animals they are intended to help, despite years of effort and substantial investment. For an industry operating with tighter margins and smaller markets than human pharma, each failure carries an even heavier cost. The process is not unlike natural selection in the wild: countless candidates compete, but only a few prove fit enough to survive. Yet just as evolution can be shaped by selective pressures, animal-health R&D can be guided by smarter, earlier decision-making.

This question lies at the heart of our work at Biorelate. By harnessing advanced AI to quantify the Probability of Technical Success (PTS) from the earliest stages of discovery, my team and I aim to bring data-driven foresight to animal-health R&D; helping innovators reduce risk, allocate resources more effectively, and ultimately deliver better medicines for animals and the systems that depend on them.

At Biorelate, we are driving what I see as a quiet but meaningful shift in how early discovery decisions are made in animal health. Rather than relying on intuition, siloed datasets, or time-intensive manual literature reviews, often spread across species and disease areas, our approach applies large-scale, AI-driven curation of biomedical and veterinary knowledge to bring greater rigor to early-stage decision-making.

At the core of this approach is an extensive knowledge graph built from millions of scientific publications and structured data sources, spanning human and animal biology. Using natural language processing and large language model (LLM)-powered algorithms, our Galactic AI platform automatically extracts cause-and-effect relationships, linking genes, targets, diseases, phenotypes, and compounds, across species at a scale and speed no human team could achieve alone.

This enables animal-health researchers to rapidly surface mechanistic insights, potential safety signals, and biomarker relationships that might otherwise require months of manual, cross-species literature exploration. Instead of searching endlessly through papers from disparate domains, scientists can focus on interpreting curated, hypothesis-relevant evidence.

This data-rich foundation fundamentally changes how we estimate a project's PTS in animal health. By illuminating the "dark knowledge" buried across both biomedical and veterinary literature, and linking it into an actionable, cross-species evidence map, we provide a more objective basis for evaluating early discovery ideas. Is a target implicated in disease-relevant pathways across one or more animal species? Have similar mechanisms failed previously due to toxicity, lack of efficacy, or species-specific biology? Are there translational biomarkers that could signal efficacy or safety early, before costly development decisions are locked in? Our curated knowledge graph is designed to answer these questions with traceable, evidence-backed insight, reducing reliance on intuition and incomplete information.

Probability of Technical Success (PTS) is a metric that captures the likelihood that a product will successfully progress through its development pathway. In animal health, this means answering a practical question early on: if we embark on this program, what are the odds it will demonstrate meaningful efficacy and an acceptable safety profile in the target species?

Historically, PTS has been calculated late in development, once candidates are already advancing through pivotal studies, often borrowing frameworks from human pharma that focus on clinical trial outcomes in people. In animal health, however, technical success hinges on additional complexities: species-specific biology, translational relevance across models, field-trial performance, and practical use in real-world settings. PTS is frequently discussed alongside Probability of Regulatory Success (PRS), which reflects the likelihood of meeting regulatory requirements for veterinary approval. Together, PTS and PRS form the composite Probability of Technical and Regulatory Success (PTRS) - the probability that a product will both work as intended and ultimately receive approval to reach veterinarians, producers, and pet owners.

As in human pharma, PTRS often dominates portfolio planning discussions in animal health because it speaks directly to a program's ultimate fate: will this innovation become an approved veterinary medicine, or will it stall before delivering value to animals and the systems that depend on them? Traditionally, estimating PTS in animal health has relied heavily on expert opinion and relatively simple models. Development teams often convene a small group of experienced veterinarians, toxicologists, or R&D leaders to review the available data (frequently sparse or uneven across species) and assign a percentage likelihood of success based largely on collective judgment. Many organizations still lean on historical benchmarks, such as average success rates by development stage, or on external expert panels to derive these probabilities.

While these approaches are preferable to proceeding without any structured assessment, they are inherently subjective and constrained. In animal health, where datasets are smaller and biological variability across species is high, expert judgment is especially vulnerable to bias, overconfidence, and incomplete visibility into prior failures or safety signals. As we have seen through our own interactions with animal-health R&D teams, PTS often hinges on judgment calls that can vary widely depending on individual experience, species familiarity, or therapeutic area. In practice, this means early-stage PTS scoring is frequently manual and rudimentary - a best-effort estimate built on partial information, rather than a comprehensive, evidence-driven view of technical risk.

This dynamic nature is precisely what makes PTS so valuable in animal health. When assessed rigorously, it becomes a forward-looking tool that informs go/no-go decisions, resource allocation, and portfolio strategy.

A high PTS supports continued investment in a program. A low PTS may prompt caution, additional studies, or early termination. Historically, however, the practical impact of PTS has been felt relatively late in development, often at major decision points such as committing to large-scale field trials or pivotal regulatory studies. By that stage, substantial time and capital have already been invested, and strategic flexibility is limited. The irony is that many of the biological and mechanistic signals that determine ultimate success are present much earlier. Wouldn't it be more effective to apply PTS-driven thinking at the very inception of an animal-health program, before resources are locked in, species choices are fixed, and avoidable risk has already been assumed?

The concept we are championing is elegantly simple: take the same risk-assessment philosophy that underpins PTRS and apply it from the very beginning of animal-health R&D. In later-stage development, PTRS helps organizations decide whether to commit substantial resources to pivotal studies or large-scale field trials. Now imagine applying analogous, data-driven probability estimates when choosing which targets to pursue, which biological hypotheses are most credible in a given species, or which compound series are truly worth optimizing.

By front-loading discovery with a probability-of-success mindset, animal-health teams can make earlier, more informed choices, dramatically improving efficiency and return on investment in an environment where development budgets are tighter and missteps are harder to reverse. It is, in effect, the application of hindsight at the outset: learning systematically from past successes and failures, captured across the literature and development history, to guide new programs before major resources are committed.

It aligns closely with the industry's need to generate sustainable returns in the face of persistent productivity challenges, long development cycles, and an increasingly selective investment environment.

Implementing this conceptual shift demands robust data and analytics, which is precisely where Biorelate comes in. By unifying decades of biomedical and veterinary knowledge into an accessible, structured platform, we make it practical to calculate and compare probabilities at the earliest stages of animal-health discovery. Researchers can query the knowledge graph to ask concrete, species-relevant questions: how often has a given target class succeeded in similar indications or animal species? What known biological mechanisms or safety liabilities have driven past failures in this pathway? The answers are grounded in quantitative risk scores, traceable evidence, and the supporting scientific literature, augmenting the intuition of scientists.

In effect, this brings the probabilistic discipline long applied in late-stage development, where statistical reasoning informs high-stakes decisions, into the earliest phases of research. The result is a shift from open-ended exploration toward a more informed, strategic approach to animal-health innovation, without constraining scientific creativity.

To be clear, this approach does not guarantee that every animal-health program will succeed. Biology remains unpredictable, species differences can surprise even experienced teams, and serendipity will always play a role in discovery. But by applying probability-of-success thinking from the very earliest stages, we can meaningfully stack the odds in our favor. It is akin to having a seasoned guide at the outset of a difficult expedition, rather than bringing one in halfway through. Early guidance can reveal which paths are likely dead ends and which hold genuine promise, helping teams avoid costly detours and irreversible commitments. In the same way, Biorelate's PTS-focused approach helps animal-health R&D teams navigate the vast search space of potential targets, mechanisms, and species strategies more intelligently, charting a course with fewer avoidable failures and a clearer line of sight to successful veterinary medicines.

This emerging paradigm, where we are using AI to predict and improve PTS from day one, signals a new era in animal-health discovery. It brings together the exploratory creativity that drives early research with the disciplined pragmatism of risk analysis, resulting in a process that is both imaginative and rigorously grounded.

We see a development landscape where more high-quality veterinary medicines reach the animals that need them, on leaner timelines, with tighter budgets, and with fewer avoidable failures along the way.

The same probabilistic thinking that once guided only late-stage go/no-go decisions can now shape ideas at their inception, improving productivity at the very beginning of the pipeline, rather than later on. This shift represents a new way of thinking about animal-health discovery. One that envisions a future where every hypothesis is accompanied by an informed probability, and every experiment moves the field closer to more predictable, sustainable success.