De-risking target identification through multidimensional data integration
Date
Sep 1, 2025
The Big Picture: Why Target Selection Is So Hard
Bringing a new drug to market often takes over a decade and more than a billion dollars. Choosing the right target is among the most decisive steps in this process, ultimately determining whether years of investment yield a viable therapy or end in failure. Yet nearly 90% of drug candidates never reach patients. Between 2010 and 2017, ~40–50% of trial failures stemmed from insufficient efficacy, and ~30% from safety issues (1,2,3)

At the root of this problem is biology’s complexity. Animal models cannot fully capture human physiology, while insights into targets, pathway redundancies, and tissue-specific liabilities are scattered across thousands of datasets and publications. Without systematic integration, warning signals are overlooked, missteps accumulate, and late-stage failures multiply. Better mining of this knowledge could reduce attrition and save billions, making integrative frameworks one of drug discovery’s most urgent needs.
Lessons from Synthetic Lethality: USP1 as a Case Study
Synthetic Lethality: From Genetics to Oncology
Synthetic lethality (SL), first described in classical genetics, occurs when simultaneous loss of two genes causes cell death. In 1997, Hartwell et al. proposed applying SL to “gene + drug” interactions (4)
, and in 2005 PARP inhibitors were shown to selectively kill BRCA1/2-deficient tumors (5,6)
. Their success made SL a leading precision oncology strategy, offering three major advantages:
Tumor Selectivity:
Selectively lethal only in cancer cells with specific alterations, sparing normal tissuesExpanded Druggable Space:
SL enables indirect targeting of tumor suppressor mutations that are otherwiseintractable (undruggable) by inhibiting their synthetic lethal partners.Precision Medicine:
Because SL depends on defined genetic contexts, biomarkers can be used to identify patients most likely to benefit, improving clinical outcomes.

Despite enthusiasm, translation has often faltered due to resistance, limited reproducibility, toxicity, and modest efficacy.
USP1 Inhibition: Promise and Risks. A Story Still Unfolding
USP1, a deubiquitinating enzyme, attracted interest in 2019 when a study reported its synthetic lethal interaction with BRCA1 deficiency (7)
. USP1 regulates multiple pathways including the Fanconi anemia pathway, DNA damage tolerance, homologous recombination, immunity (TBK1), survival (AKT), differentiation (C/EBPβ), and stemness (ID proteins). Furthermore, USP1 inhibition can induce chronic PCNA monoubiquitination and genomic instability in BRCA-wild-type cells, raising not only toxicity concerns but also uncertainty about which biomarkers will predict clinical response (8)
.
Box. USP1 inhibition: Clinical Setbacks and Ongoing Efforts
Compound | Developer | Current Status | Key Notes |
---|---|---|---|
RO7623066 | Roche / KSQ | Early clinical | Manageable anemia; strong PD signal; limited efficacy (1 PR in 42 patients) |
TNG348 | Tango | Discontinued | Stopped due to Grade 3/4 liver enzyme elevations |
SIM0501 | Simcere | Early clinical | Trial ongoing |
HSK39775 | Haisco | Early clinical | Trial ongoing |
Debio 0432 | Debiopharm | Late preclinical | Preclinical |
XL309 | Exelixis (licensed from Insilico) | Early clinical | Developed by Insilico Medicine using AI and licensed to Exelixis in 2023; early clinical stage |
While the readouts for additional molecules will provide a definitive picture and current efforts might benefit from recent setbacks, emerging human data have shown both modest antitumor activity and notable toxicities. These setbacks highlight how even compelling synthetic lethality targets like USP1 may face translational challenges, underscoring the critical need for deeper, integrative evaluation frameworks to better predict clinical success.
An Integrative Frameworks to De-Risk Target Identification (ID)
De-risking target ID requires integrating diverse data streams including chemistry, metabolism, pharmacokinetics, and mechanistic biology. A unified workflow that connects these layers can reveal liabilities and mechanistic risks early, well before clinical testing. Key components of this approach must include:
Target-Level Risk Profiling
Aggregate multi-omic expression data across tissues and species to reveal on-target safety liabilities. Consolidate genetic model and functional genomic data to uncover efficacy and toxicity signals. Map pathway redundancies to anticipate compensatory mechanisms that may undermine target vulnerability.
De-Risk Chemistry and PK/PD Early
Evaluate structural tox alerts, metabolic soft spots, and reactive metabolite formation to inform scaffold selection early in discovery. Model ADME properties and exposure–response relationships to prioritize candidates with the best balance between efficacy and safety. Recognizing when chemotype-related liabilities may limit the therapeutic index is especially important in clinical settings where patients receive longer treatments than modeled preclinically. These insights promote informed go/no-go decisions and guide resource allocation across preclinical and clinical development, as well as between competing programs. While several tools exist, they remain fragmented and poorly integrated into streamlined platforms.
Bridging Clinical Translation Gaps
Align preclinical models with human genetic, transcriptomic, and clinical datasets to refine hypotheses and patient selection. Integrate interspecies PK scaling, mechanistic PK/PD models, and population variability simulations for accurate first-in-human dose prediction. Map preclinical biomarkers to clinical surrogates to improve regulatory readiness. Account for confounders such as genetic polymorphisms or concomitant medications to strengthen trial design.
Closing Thoughts
The USP1 case study illustrates how promising biological rationales, such as synthetic lethality, can encounter sharp clinical limitations. Reducing this gap requires integrative, multidisciplinary frameworks that bridge biology, chemistry, and clinical translation.
Developing comprehensive, multimodal data integration strategies will be essential for early de-risking and more successful decision-making. With the support of modern computational approaches, including artificial intelligence, this level of multidimensional integration is now feasible at scale, enabling researchers to make faster and better-informed choices. The future of drug discovery will not be defined by single datasets, but by how effectively we integrate them, accelerating the path from target discovery to patient benefit.
References
Hinkson, I. V., Madej, B. & Stahlberg, E. A. Accelerating Therapeutics for Opportunities in Medicine: A paradigm shift in drug discovery. Front. Pharmacol. 11, 770 (2020).
Harrison, R. K. Phase II and phase III failures: 2013–2015. Nat. Rev. Drug Discov. 15, 817–818 (2016).
Dowden, H. & Munro, J. Trends in clinical success rates and therapeutic focus. Nat. Rev. Drug Discov. 18, 495–496 (2019).
Hartwell, L. H., Szankasi, P., Roberts, C. J., Murray, A. W. & Friend, S. H. Integrating genetic approaches into the discovery of anticancer drugs. Science 278, 1064–1068 (1997).
Bryant, H. E. et al. Specific killing of BRCA2-deficient tumours with inhibitors of poly(ADP-ribose) polymerase. Nature 434, 913–917 (2005).
Farmer, H. et al. Targeting the DNA repair defect in BRCA mutant cells as a therapeutic strategy. Nature 434, 917–921 (2005).
Lim, K. S. et al. USP1 is required for replication fork protection in BRCA1-deficient tumors. Mol. Cell 72, 925–941.e4 (2018).
Simoneau, A. et al. Ubiquitinated PCNA drives USP1 synthetic lethality in cancer. Mol. Cancer Ther. 22, 215–226 (2023).