What stands between a promising biomarker and a precision medicine breakthrough?

Precision medicine innovators have access to more biological data and more powerful ways to interrogate it than ever before. Genomics, transcriptomics, proteomics and other high-dimensional technologies can reveal increasingly complex molecular signatures, while artificial intelligence (AI) and machine learning can uncover patterns across these data at unprecedented scale.

But finding a signal is not the same as knowing what it means for drug development.

For companies developing precision therapeutics, the critical question is increasingly: Which biological signals identify clinically meaningful patient subtypes, predict treatment response or point to the patients most likely to benefit from a therapy?

Answering that question is difficult when much of the available molecular evidence captures only a moment in a patient’s disease journey. A molecular profile may reveal what is detectable in a sample, but without knowing what came before and what happened afterward, it can be difficult to determine whether a signal reflects an underlying disease subtype, current disease activity, treatment exposure or something predictive of future response.

That challenge extends across oncology, neuroscience, rare disease, cell and gene therapy, autoimmune and inflammatory disease, and other areas where patient heterogeneity can shape therapeutic outcomes.

For precision medicine developers, the challenge is understanding what molecular data means in the context of a patient’s disease course, treatment exposure and response over time.

Longitudinal clinical context can provide that perspective, enabling researchers to assess when and how a molecular signal may be relevant to drug development.

Why molecular signals need clinical context

The ambition of precision medicine is not simply to characterize disease at greater molecular resolution. It is to translate that resolution into better development decisions: identifying the right target, defining the right patient population, developing the right biomarker and ultimately matching therapies to the patients most likely to benefit.

This becomes difficult when heterogeneous diseases are treated as single clinical categories.
Two patients with the same diagnosis may have very different underlying biology, rates of progression, treatment histories and responses to therapy. A biomarker measured at one moment may have very different significance depending on where each patient is in that trajectory.

The challenge cuts across therapeutic areas, although it takes different forms. Oncology has demonstrated the potential of molecularly defined patient populations, but many diseases do not have an equivalent of the tumor as a central molecular reference point. In neuroscience, rare disease, immunology and other complex conditions, relevant biology may be distributed across tissues and systems and change over time.

The implication for precision drug development is straightforward. The biological signal is only part of the evidence, and its value depends on understanding the clinical trajectory surrounding it.

What cross-sectional data can miss in dynamic disease

Autoimmune and inflammatory diseases illustrate the importance of longitudinal context particularly well. Patients sharing a diagnosis such as rheumatoid arthritis or inflammatory bowel disease can differ substantially in disease activity, treatment history, flare patterns and long-term therapeutic response. Relevant biology may also change as disease progresses or as a patient responds, fails to respond or loses response to treatment.

A cross-sectional biospecimen captures one moment in that dynamic process.
Consider a promising molecular signature identified in a patient with rheumatoid arthritis. Its significance depends on questions that the sample alone cannot answer.

• Was it collected before or after treatment?
• Was the patient experiencing a flare?
• Did disease activity subsequently improve?
• Was that improvement sustained?
• Did the patient eventually lose response or change therapy?

These distinctions can determine whether a molecular signal is simply interesting or potentially actionable.

Treatment response illustrates the problem. A patient who never responds to therapy is clinically different from one who responds initially and loses benefit months later. If those trajectories reflect different underlying biology, reducing both to a simple responder or nonresponder classification can obscure precisely the information precision medicine is intended to uncover.

Start with the development question, not the data

Addressing this challenge requires thinking differently about how precision medicine evidence is generated.

Rather than beginning with the largest available molecular data set, developers can begin with the decision they need to make and work backward.

• What biological question needs to be answered?
• Which patients are most informative?
• At what points in the disease or treatment journey should biology be measured?
• What clinical outcomes are needed to interpret those measurements?

One approach is to connect longitudinal clinical data with biospecimens collected at meaningful points in the patient journey. PPD™ CorEvitas™ Clinical Registries apply this model through clinician-curated longitudinal data and linked biorepositories, allowing molecular findings to be evaluated in the context of treatment and outcomes over time.

The value of this approach is the ability to interpret what a biological signal may mean clinically. Genomic, transcriptomic, proteomic and other molecular findings can be examined in relation to disease state, treatment exposure and what happens to the patient next. That can help researchers distinguish a signal associated with an underlying disease subtype from one reflecting treatment effect or a transient disease state and help identify which signals warrant further investigation as potential predictors of response or durability.

The goal is evidence that can translate biological insights into more informed development decisions.

Moving AI from pattern recognition to clinical relevance

This need for clinically grounded evidence becomes even more important as AI-driven drug discovery expands.

AI and machine-learning models can identify relationships across thousands of clinical and molecular variables, potentially revealing previously unrecognized patient subtypes, pathway activation patterns and candidate biomarkers. But computational sophistication cannot compensate for data that lack meaningful clinical context.

For drug developers, the opportunity is to use AI not simply to identify patterns, but to determine which patterns may have clinical and developmental relevance.

Purposefully timed biospecimens linked to longitudinal clinical data can enable computationally identified signals to be evaluated in relation to treatment initiation, early response, durable benefit, flare and loss of response. This helps move AI-generated findings from statistical associations toward clinically grounded hypotheses that can be tested and independently validated.

For AI-health companies, biotech organizations and emerging precision medicine developers, this distinction matters. The value of an AI model ultimately depends not only on its sophistication, but on the quality and clinical context of the evidence it is built to interrogate.

Turning molecular insight into better development decisions

The value of molecular insight ultimately lies in how it informs drug development.
Placing molecular signals within clinically meaningful patient trajectories can enable developers to assess which patient subtypes to pursue and which biomarkers may have practical development value.

These insights can inform consequential program decisions.

• Which population is most appropriate to study?
• Could a biomarker support enrichment?
• Which endpoints may best capture benefit?
• Does a mechanism appear more compelling in one disease or subgroup than another?
• Which findings warrant prospective validation?

Oncology has demonstrated the potential of using molecular biology to define patient populations. As precision medicine expands across other heterogeneous conditions, applying this approach will require deeper understanding of how biology relates to disease course, treatment and outcomes over time.
Longitudinal clinical context provides a way to make those connections. It can help researchers move beyond what is detectable in a sample to understand when a biological signal may be meaningful, for whom and under what clinical circumstances.

As molecular technologies and analytical capabilities continue to advance, their impact on drug development will depend on our ability to put the signals they uncover into meaningful clinical context. Doing so brings precision medicine closer to answering the question at the heart of the field: which therapy is most likely to benefit which patient, and why?