The genomics revolution in oncology has been genuinely productive. Pan-tumor sequencing programs have catalogued somatic alterations at a scale that was unimaginable a decade ago. Large public cohort datasets now cover tens of thousands of tumor samples across dozens of histologies, with alteration frequencies, co-occurrence patterns, and expression context available for nearly every gene of interest. For a discovery team trying to identify where to focus, this is valuable starting material.
It is also, by itself, insufficient. The distance between a genomically observed alteration and a target with enough evidence to justify first-assay commitment is larger than it often looks at the nomination stage. Closing that gap requires evidence that most genomic datasets do not provide. This post is about where that evidence gap lies and how to assess it systematically.
Frequency Is Not Mechanistic Relevance
The most common starting point for genomic target nomination is alteration frequency: a gene that is mutated, amplified, or deleted at measurable frequency in the tumor type of interest is a candidate worth examining. The reasoning is intuitive and sometimes valid. If a gene is altered in 20% of a given tumor type, and if those alterations are non-random in their pattern, there is a reasonable prior that the gene has functional relevance in that tumor biology.
The problem is that frequency is a poor proxy for functional significance in a meaningful fraction of cases. Driver versus passenger status in somatic alteration is not determined by frequency alone. Some low-frequency alterations are strong functional drivers in specific tumor subtypes. Some high-frequency alterations are passengers that occur in genomically unstable regions. The genomic datasets can show you the frequency. They cannot show you whether the alteration is causing the tumor phenotype or merely occurring alongside it.
Moving from observed frequency to mechanistic relevance requires a different evidence type: experimental data showing that the alteration actually changes cellular behavior in a tumor-relevant way. This is the step where genomic nomination hands off to biological validation, and it is the step that takes the most time and generates the most failures.
The Functional Validation Layer
Functional validation asks whether manipulating the target, by knockout, knockdown, overexpression, or pharmacologic inhibition, produces the expected phenotypic effect in a relevant tumor cell model. This is the minimum requirement for moving a genomically nominated target toward confidence. Without it, you have an association. With it, you have a directional biological signal.
The quality of functional validation varies considerably. A loss-of-function study in a two-dimensional cell culture system, using a commercially available cell line that may or may not carry the relevant alteration, provides a weak signal. The same experiment repeated across multiple independent cell lines with confirmed alteration status, ideally followed by an in-vivo validation in a genetically engineered or patient-derived xenograft model, provides substantially stronger evidence.
The challenge in a time-constrained synthesis is that both types of study are present in the literature, and both are cited in target reviews. Without classifying the functional validation evidence by experimental system and independence of replication, the difference between weak and strong functional support gets obscured. A target with twelve cell-line studies and no in-vivo data can appear more validated than a target with two well-designed in-vivo studies in the relevant tumor context. The paper count says twelve versus two. The evidence quality picture says the reverse.
Clinical Context as the Second Validation Layer
Clinical context evidence provides a distinct and necessary dimension beyond preclinical functional validation. The question this evidence addresses is not whether the target is biologically functional in a cell or mouse model, but whether there is any signal in human tumor data or clinical context that the target is relevant to patient biology.
This can come from multiple sources. Expression data from tumor versus normal tissue from large tumor genomics programs provides a basic layer: is the target differentially expressed in the tumor compartment in a direction that is consistent with the mechanistic hypothesis? Alterations in the target that are associated with clinical outcomes, even in observational studies, provide a stronger clinical context signal. The presence of the target in the therapeutic mechanism of agents that have shown any clinical activity, even in early-phase trials in a relevant indication, provides contextual alignment that is meaningful even when direct clinical target data is sparse.
The clinical context layer does not replace preclinical functional validation. These are orthogonal evidence dimensions, and a strong preclinical case without any clinical context support is a different risk profile than a case where the two align. Both dimensions contribute to the composite confidence picture, and divergence between them is as informative as alignment.
Where Programs Most Often Lose Time
In our experience working with oncology research programs, the evidence gap between genomic nomination and first-assay commitment is most often underestimated in two specific situations. The first is when genomic frequency is high but functional validation is thin. Teams sometimes enter assay development with strong genomic justification but limited functional support, reasoning that the frequency data is compelling and functional validation can be done in parallel. This sometimes works. It also sometimes produces assay results that are not consistent with the genomic hypothesis, requiring a return to validation before the target can be advanced.
The second situation is when the functional validation evidence is largely from a single experimental context, usually one cell-line system, and has not been tested in a more tumor-relevant model. Single-context functional evidence is vulnerable to the known limitations of in-vitro systems: the absence of stromal interactions, immune components, and the physical constraints of the tumor microenvironment that can fundamentally change target biology. A target that shows strong effects in one cell line but has never been tested in anything more complex carries more uncertainty than the in-vitro evidence suggests.
What a Systematic Evidence Assessment Adds
A structured evidence synthesis that classifies validation evidence by type and quality tier surfaces these gaps explicitly. Rather than a count of papers or a list of citations, the output shows where the evidence is strong (in-vivo, replicated, tumor-type-specific functional validation with consistent clinical context alignment) and where it is thin (genomic frequency without functional corroboration, single-context in-vitro data, no clinical context alignment).
That distinction is most valuable when a discovery team is comparing targets at the nomination stage, before any bench resources commit. A ranking that reflects evidence quality, not just volume, changes which targets a team prioritizes for first-assay investment. It does not guarantee that the top-ranked target will succeed. Assay development introduces new variables that no literature synthesis can fully anticipate. But it systematically reduces the frequency of situations where evidence quality was the reason for failure.
We are not claiming the gap between genomic observation and druggable target is fully closeable through literature synthesis. Truly novel target classes, especially those in underexplored tumor types, may have thin evidence bases that accurately reflect the actual state of the science. A systematic synthesis cannot manufacture validation that does not exist. What it can do is make the actual state of the evidence clear before commitment, rather than after.
If you are working through a genomics-to-target pipeline and want to understand what the evidence picture looks like for your current candidates beyond the genomic data layer, the Avenzo platform is built for that analysis. Reach out to discuss what a structured evidence assessment would look like for your program.