In early-stage oncology discovery, the decision that determines where bench time goes is target selection. Everything downstream, the assay development, the compound screening, the mechanistic follow-up, runs on the premise that the target is worth pursuing. When that premise is wrong, the bench time spent is not just wasted. It is borrowed from work that could have been done on a better candidate. The cost is not just the failed experiment. It is the opportunity never taken.
We think about this a lot at Avenzo, because the product we are building is fundamentally a tool for making that initial premise better-supported. Understanding the real economics of target selection is part of why we built it the way we did.
How Resources Accumulate Before the First Clear Signal
Target nomination and the early investigation phase that follows account for a significant fraction of discovery resource consumption in a typical oncology program. Before a team reaches a clear mechanistic decision point, a candidate will typically have consumed: literature review and evidence assembly time (often weeks to months of scientist effort), initial target characterization assays, cell-line panels to understand alteration context and baseline sensitivity, reagent development for target engagement confirmation, and sometimes early hit identification work before the target hypothesis is fully tested.
The failure signal often does not arrive until this work is well underway. In mechanistic oncology targets, the common failure pattern is not an absence of signal in the first assay. It is a signal that looks encouraging in the initial screen, weakens in the next model system, and then fails to hold up when the biology is examined in a more relevant context. By the time the team recognizes that the target hypothesis was not supported by what the evidence actually said, the accumulated investment is substantial.
We do not have a precise universal number for this, and anyone who gives you one is extrapolating from limited data. What we observe in conversations with discovery teams is that the investigation phase per target candidate, from nomination to clear go/no-go, is measured in months and involves multiple scientists. For a discovery team of four to eight people working on a defined therapeutic area, the opportunity cost of an investigation cycle on a target that ultimately fails at the mechanistic level is real in terms of program timeline.
Where the Evidence Gap Actually Lives
When a target fails early in investigation after passing nomination, it rarely fails because the team made a reasoning error with the evidence they had. It usually fails because the evidence used to make the nomination decision was thinner than it appeared. A target that has twenty publications does not necessarily have strong mechanistic support if most of those publications are from a single research group, a single cell-line system, and concentrate in in-vitro rather than in-vivo experiments. The volume looks like confidence. The quality distribution tells a different story.
This gap between apparent evidence richness and actual evidence quality is the central problem we designed the Avenzo framework to address. The quality tier structure is not an academic classification exercise. It is a tool for making visible whether the evidence behind a target nomination would actually support the early investigation decision, or whether it is one good paper surrounded by lower-weight confirmation.
A practical way to describe the problem: if you ask most discovery teams "what's the evidence for Target X," you will get an accurate description of the most prominent papers supporting it. You will rarely get a comprehensive accounting of how many independent experimental systems have tested it, whether in-vivo replication exists from a second group, and whether there are any published null or negative results that did not make it into the standard narrative around the target. The literature synthesis that most teams have capacity to do is favorable toward the targets they are already considering. That is a structural bias in how target selection happens, not a flaw in any individual team's process.
The Asymmetry of Sunk Cost and Opportunity Cost
One of the harder dynamics to manage in discovery programs is that the sunk cost of investigation work on a current target is always visible and quantifiable, while the opportunity cost of the better target not being investigated is hypothetical. The team knows how much they have invested in Target X. They do not know how Target Y, which scored higher on evidence quality but was ranked lower on the team's internal assessment, would have performed in the same investigation cycle.
This asymmetry tends to extend investigation timelines past the point that evidence quality would justify. The decision to continue investigating a target that has produced ambiguous initial results is easier to rationalize than the decision to cut it and start over on a candidate with better evidence underpinning. Even when the team intellectually agrees that the current target is not holding up, the activation energy required to formally abandon the work and redirect resources is high.
The straightforward implication is that the evidence quality bar at nomination is more important than it might seem in the moment of the decision. A nomination criterion that allows weaker evidence through because the target is interesting or strategically appealing has downstream costs that are distributed across the investigation phase and rarely accounted for as evidence-quality-related expenditure.
What Changes When Nomination Evidence Is Stronger
Higher evidence quality at nomination does not guarantee that the target will succeed in early investigation. In-vivo functional evidence can be wrong. Clinical context signals can be confounded by patient population heterogeneity that did not appear in the published data. Pathway architecture can produce resistance mechanisms that are not predictable from static network analysis. We are not saying that stronger nomination evidence eliminates mechanistic failures.
What it does is change the distribution of where failures occur and why. Targets nominated with strong multi-system functional evidence and clinical context corroboration fail for different reasons than targets nominated primarily on genomic frequency and a few cell-line studies. The former category's failures are more likely to be genuinely novel biology that could not have been predicted. The latter category's failures often trace back to the evidence that was available but not adequately weighted before commitment.
From a resource allocation perspective, the difference matters. If your investigation failures are concentrated in the genuinely-unpredictable category, that is the irreducible cost of doing real discovery work. If they are concentrated in the better-evidence-would-have-caught-this category, that is a fixable cost. Separating those two categories requires being honest about what the nomination evidence actually showed, which requires knowing what it showed in the first place.
The Practical Argument for Front-Loading Evidence Quality
The economics we have described above form the practical case for investing time in evidence quality assessment before nomination, not during investigation. The unit of investment at the evidence synthesis stage is days of analyst time and computational processing. The unit of investment at the investigation stage is months of wet-lab scientist time, reagent cost, and assay development overhead.
When we talk to early-stage discovery teams about the Avenzo platform, the conversation often starts with the question: "how long does an evidence assessment take?" The implied comparison is to the effort of doing it manually. The more useful comparison is to the investigation cycle the assessment is being asked to inform. An evidence quality assessment that takes a few days of platform time, and surfaces the information that a highly-cited target has shallow in-vivo functional support from a narrow experimental base, has very different return on investment than it would if the same information emerged six months into an investigation program.
We built the platform to operate at nomination speed, not review-paper speed. The evidence is precomputed and continuously updated rather than assembled from scratch at query time. For a discovery team making target nominations quarterly or around pipeline meetings, this means the quality signal is available when the decision is being made, not after it has already been taken.
If your team is working through a list of candidates where the evidence picture feels uncertain and you want to understand what the quality distribution actually looks like across your nominees before committing investigation resources, reach out to discuss how the Avenzo evidence analysis applies to your program context.