OnCo
Cancers & treatments

Open questions

57 unresolved questions across 25 targets and 24 technologies. Each says why it is still open, what evidence would settle it, who has to act, and which trials, products and ideas in the corpus bear on it. Cancers carry their own open problems on their pages; this index covers the molecules and the methods.

57 questions
Have AI-discovered oncology molecules improved clinical success rates, or only shortened the preclinical timeline?
Technology: AI-driven drug & target discovery
Several AI-derived molecules (HER2, CDK2/4, KRAS binders) are in phase 1 and structure models are now routine, but no AI-originated cancer drug has yet reported a positive randomised trial.
AI-driven drug & target discovery
Will fourth-generation ALK inhibitors extend the more than five-year progression-free survival seen with lorlatinib, or is lorlatinib the ceiling for on-target therapy?
Target: ALK
CROWN's five-year data made lorlatinib the first-line standard; resistance now runs through compound mutations and bypass pathways that neladalkib targets, but the incremental benefit after lorlatinib is unknown.
ALK
Can off-the-shelf allogeneic CAR-T or CAR-NK products persist long enough to match autologous durability?
Technology: Allogeneic (off-the-shelf) cell therapy
Allogeneic products avoid manufacturing delay but are rejected within weeks unless the host is heavily lymphodepleted; response rates are comparable, durability is not.
Allogeneic (off-the-shelf) cell therapy
What determines payload resistance to topoisomerase-I ADCs, and can it be measured before choosing the next ADC?
Technology: Antibody-drug conjugate (ADC)
TOP1 mutations, SLFN11 silencing and efflux pumps are described, but there is no clinical assay, and sequencing two TOP1 ADCs is now common practice without one.
Antibody-drug conjugate (ADC)
Why do deruxtecan ADCs cause interstitial lung disease, and can it be predicted or prevented?
Technology: Antibody-drug conjugate (ADC)
ILD is the dose-limiting and occasionally fatal toxicity across HER2-, TROP2- and HER3-directed DXd ADCs, occurs in patients whose tumours do not express the target in lung, and has no validated predictor.
Antibody-drug conjugate (ADC)
Does B7-H3 expression predict benefit from ifinatamab deruxtecan in small-cell lung cancer and prostate cancer, or is it another ADC given without a biomarker?
Target: B7-H3
B7-H3 is broadly expressed with low normal-tissue expression, and early single-arm data are promising, but no threshold has been validated and payload resistance from prior topotecan or lurbinectedin is plausible.
B7-H3
Is fixed-duration venetoclax-based therapy curative for a subset of CLL, and who needs retreatment?
Target: BCL-2
CLL14 and AMPLIFY show long treatment-free remissions after one or two years; measurable-residual-disease kinetics predict relapse but retreatment strategies and the role of newer BCL2 inhibitors are unsettled.
BCL-2
In what order should BCMA CAR-T, BCMA bispecifics and GPRC5D-directed therapy be given, and does antigen loss after one preclude the next?
Target: BCMA
CARTITUDE-4 and MajesTEC data cover single lines; biallelic TNFRSF17 loss after bispecifics is described and T-cell fitness falls with each line, so sequencing changes outcomes in ways no trial has compared.
BCMA
Does hitting two antigens with one ADC add efficacy or only widen the eligible population?
Technology: Bispecific ADC
Izalontamab brengitecan (EGFR x HER3) shows high response rates in EGFR-mutant lung and oesophageal cancer, but no trial isolates the contribution of the second arm from payload potency.
Bispecific ADC
Does BREAKWATER's first-line encorafenib plus cetuximab plus chemotherapy improve overall survival enough to displace chemotherapy plus bevacizumab for every BRAF V600E colorectal patient?
Target: BRAF
BREAKWATER showed higher response and progression-free survival; overall survival maturity, the role of the chemotherapy backbone, and MSI-high BRAF patients (who may do better with immunotherapy) remain open.
BRAF
Should BRCA reversion mutations detected in ctDNA change treatment before radiographic progression on a PARP inhibitor?
Target: BRCA1 / BRCA2 (HRD)
Reversions restore homologous recombination and predict PARP-inhibitor and platinum failure, and they are detectable in plasma, but no trial has switched therapy on that signal alone.
BRCA1 / BRCA2 (HRD)
Do BTK degraders outperform non-covalent inhibitors once C481S, T474I or L528W mutations have appeared, and could they replace inhibitors in front line?
Target: BTK (Bruton tyrosine kinase)
Pirtobrutinib is active against C481S but selects kinase-dead L528W; degraders such as BGB-16673 remove the scaffold and are active preclinically against every known mutation, with early clinical responses.
BTK (Bruton tyrosine kinase)
Can point-of-care or non-viral manufacturing bring autologous CAR-T below the cost of a year of bispecific antibody therapy?
Technology: CAR-T cell therapy
List prices and vein-to-vein times keep CAR-T from most eligible patients worldwide; academic point-of-care programmes (Spain, India, Brazil) report far lower costs but are outside the regulatory mainstream.
CAR-T cell therapy
What is missing for CAR-T to work in solid tumours: the antigen, trafficking, persistence in a suppressive microenvironment, or all three?
Technology: CAR-T cell therapy
Approvals remain confined to B-cell and plasma-cell malignancies; solid-tumour responses (GD2 in neuroblastoma, CLDN18.2 in gastric cancer) are real but rarely durable.
CAR-T cell therapy
Can in vivo CAR-T (lentiviral or mRNA delivery of the CAR to the patient's own T cells) match ex vivo products in B-cell malignancies?
Target: CD19
Manufacturing cost, slot limits and vein-to-vein time are the main barriers to CD19 CAR-T access; in vivo approaches remove them but first-in-human data are only now emerging.
CD19
After progression on a CDK4/6 inhibitor, does continuing CDK4/6 inhibition with a new endocrine partner beat switching to a targeted or antibody-drug conjugate approach?
Target: CDK4/6
postMONARCH showed modest benefit from continuing abemaciclib; resistance runs through RB1 loss, cyclin E and PI3K activation, and genotype-directed options (capivasertib, inavolisib, oral SERDs) compete for the same slot.
CDK4/6
Should CLDN18.2 be attacked with an antibody plus chemotherapy, an ADC, or CAR-T, and at what expression threshold?
Target: Claudin 18.2
Zolbetuximab needs at least 75% moderate-to-strong staining; ADCs such as CMG901 and the Chinese-approved satri-cel appear active at lower expression, but no trial compares modalities or thresholds.
Claudin 18.2
Does comprehensive genomic profiling for every advanced-cancer patient improve survival at the population level, and who pays for the negative results?
Technology: Comprehensive genomic profiling
Guidelines recommend broad panels in lung and other cancers; uptake is uneven, most reports find no actionable driver, and randomised evidence of survival benefit from profiling itself is thin.
Comprehensive genomic profiling
Can genome-wide dependency maps in cell lines be made to predict dependencies in patients' tumours, where the microenvironment and immune system are present?
Technology: CRISPR functional genomics
DepMap has screened more than 1,000 lines and found many context-specific dependencies, but in vivo and organoid screens often disagree with plastic dishes, and few hits have reached the clinic.
CRISPR functional genomics
Does adding tarlatamab to first-line chemo-immunotherapy in small-cell lung cancer improve survival, and can DLL3 expression or subtype select patients?
Target: DLL3
DeLLphi-304 proved second-line benefit; DeLLphi-305 tests first-line maintenance with durvalumab. DLL3 is not tested before treatment, yet expression varies by ASCL1 subtype and falls with plasticity.
DLL3
Can two payloads on one antibody be dosed to full effect, or does combined toxicity force each below its active dose?
Technology: Dual-payload ADC
Dual-payload ADCs aim to pre-empt payload resistance; preclinical data are encouraging but the first clinical programmes must show that the therapeutic window is not simply halved.
Dual-payload ADC
Which agent should follow osimertinib when C797S emerges: a fourth-generation TKI, amivantamab-based therapy, or an ADC that ignores genotype?
Target: EGFR
C797S removes the covalent anchor of every third-generation TKI. Allosteric fourth-generation inhibitors exist only in early trials, while ADCs and bispecifics work regardless of the mutation but carry different toxicities.
EGFR
Which patients need intensified first-line therapy (osimertinib plus chemotherapy, or amivantamab plus lazertinib) and which are well served by osimertinib alone?
Target: EGFR
FLAURA2 and MARIPOSA improve progression-free survival at the price of toxicity, and subgroup signals (brain metastases, TP53 co-mutation, detectable ctDNA) suggest the benefit is not uniform.
EGFR
Does switching to an oral SERD when ESR1 mutations first appear in ctDNA, before progression, improve overall survival?
Target: Estrogen receptor (ERα)
SERENA-6 showed longer progression-free survival with an early camizestrant switch, but overall survival and quality-of-life benefit are unproven and the strategy requires serial ctDNA testing.
Estrogen receptor (ERα)
Can FGFR2-selective inhibitors overcome the polyclonal kinase-domain mutations that end responses to pemigatinib and futibatinib?
Target: FGFR2
Resistance in cholangiocarcinoma is polyclonal (N550, V565, E566, L618) and varies between lesions; covalent and next-generation FGFR2-selective agents cover subsets, and hyperphosphataemia from FGFR1 limits dosing of pan-FGFR drugs.
FGFR2
Does the FLASH effect (normal-tissue sparing at ultra-high dose rate) hold in people at curative doses?
Technology: FLASH radiotherapy
The effect is robust in animals; first-in-human studies used palliative bone metastases with electrons or protons, and the mechanism (oxygen depletion, radical recombination) is unsettled.
FLASH radiotherapy
Can folate-receptor-alpha ADCs work below the 75% high-expression cut-off used for mirvetuximab, and in platinum-sensitive disease?
Target: Folate receptor alpha
MIRASOL enrolled only FRalpha-high platinum-resistant patients; newer FRalpha ADCs with topoisomerase payloads (rinatabart sesutecan, luveltamab tazevibulin) claim activity at lower expression but lack randomised data.
Folate receptor alpha
After trastuzumab deruxtecan, does a second HER2-directed ADC with a different payload work, or is payload resistance shared?
Target: HER2
Resistance to topoisomerase-I payloads (TOP1 mutation, SLFN11 loss, efflux) is target-independent, so a second TOP1 ADC often fails; whether switching payload class restores benefit has not been tested prospectively.
HER2
How should pathologists reliably separate HER2-low and HER2-ultralow from HER2-zero when the assays were designed to find HER2-high?
Target: HER2
DESTINY-Breast04 and 06 made IHC 1+ and faint IHC 0 staining actionable, yet inter-observer agreement at the low end is poor and the antibodies, controls and cut-offs were validated for amplification.
HER2
Should homologous-recombination deficiency be measured as a genomic scar (past history) or as current repair function?
Technology: HRD & BRCA testing
Scar assays (myChoice GIS, HRDetect) predicted PARP-inhibitor benefit in PAOLA-1 but stay positive after resistance develops; functional assays (RAD51 foci) may track current status but are not standardised.
HRD & BRCA testing
When should vorasidenib be started in IDH-mutant glioma, and can it delay or replace radiotherapy and chemotherapy?
Target: IDH1 / IDH2
INDIGO enrolled grade 2 patients after surgery only; whether IDH inhibition helps after radiotherapy, in grade 3 disease, or as a way to defer radiotherapy-related cognitive harm is untested.
IDH1 / IDH2
How should regulators assess integration, off-target transduction and germline risk for in vivo CAR delivery vectors?
Technology: In vivo CAR-T
In vivo lentiviral and LNP-mRNA CAR approaches transduce cells inside the patient, so the safety framework built for ex vivo manufacturing (release testing, vector copy number per product) does not apply directly.
In vivo CAR-T
Can shared-antigen KRAS vaccines prevent relapse after surgery in pancreatic and colorectal cancer?
Target: KRAS
Mutant KRAS peptides are public neoantigens present in most pancreatic cancers, but AMPLIFY-7P missed its endpoint and it is unclear whether T-cell responses translate into fewer relapses.
KRAS
Does blocking the adaptive receptor feedback up front (EGFR, SHP2 or SOS1 co-inhibition) beat sequencing after progression?
Target: KRAS
Relief of ERK feedback re-activates receptors within hours of RAS inhibition; CodeBreaK 300 showed EGFR co-blockade helps in colorectal cancer, but whether that generalises to lung cancer and to SHP2 or SOS1 combinations is open.
KRAS
Will KRAS G12D and pan-RAS(ON) inhibitors improve survival in pancreatic cancer, where G12D dominates?
Target: KRAS
The covalent G12C chemistry does not transfer to G12D, and pancreatic tumours rewire quickly through receptor feedback. Daraxonrasib reached approval on RASolute 302 but durable benefit in first line and in the adjuvant setting is unproven.
KRAS
How should ctDNA assays be standardised so that a result from one laboratory means the same as another's?
Technology: Liquid biopsy (ctDNA)
Tumour-informed and tumour-naive assays differ in sensitivity by an order of magnitude; reference materials and reporting standards are immature, and trial results are assay-specific.
Liquid biopsy (ctDNA)
Do menin inhibitors added to induction chemotherapy or venetoclax-azacitidine improve survival in newly diagnosed NPM1-mutant and KMT2A-rearranged AML?
Target: Menin
Revumenib and ziftomenib were approved on single-arm relapsed data; MEN1 resistance mutations and differentiation syndrome are described, and front-line randomised trials are only now under way.
Menin
What definition of MET amplification or overexpression should select patients for MET-directed drugs at EGFR-TKI progression?
Target: MET
Copy-number thresholds by FISH, NGS and IHC disagree; MARIPOSA showed amivantamab's benefit without MET selection, while MET TKI combinations depend on it.
MET
Does treating ctDNA-detected molecular residual disease improve survival, or only move the diagnosis of relapse earlier?
Technology: MRD / molecular residual disease testing
DYNAMIC showed ctDNA can safely de-escalate adjuvant chemotherapy and IMvigor011 showed benefit of atezolizumab in ctDNA-positive bladder cancer, but escalation on a positive result has not improved survival in most cancers yet.
MRD / molecular residual disease testing
Do multi-cancer early detection blood tests reduce cancer deaths, and at what cost in false positives and workups?
Technology: Multi-cancer early detection (MCED)
PATHFINDER showed feasibility and NHS-Galleri is the first randomised trial, but stage shift is not proof of mortality benefit, and positive results need a resolution pathway.
Multi-cancer early detection (MCED)
Can oncolytic viruses be delivered systemically and still reach tumours, or will they remain intratumoural agents?
Technology: Oncolytic viruses
Approved and late-stage products (T-VEC, cretostimogene, RP1) are injected into accessible lesions; neutralising antibodies and hepatic clearance limit intravenous use.
Oncolytic viruses
Do pathology foundation models improve patient outcomes prospectively, or only benchmark scores on retrospective slides?
Technology: Pathology & radiology foundation models
Dozens of models report state-of-the-art on TCGA-derived tasks; almost none have prospective, multi-site validation with clinical endpoints, and scanner and site batch effects remain.
Pathology & radiology foundation models
Can patient-derived organoid drug testing predict a person's response well enough to guide therapy, and fast enough to matter?
Technology: Patient-derived organoids
Retrospective concordance is high in colorectal and pancreatic cancer, but organoids take weeks, fail to grow in a third of cases, lack immune and stromal components, and no randomised trial has shown benefit.
Patient-derived organoids
For which cancers should neoadjuvant immunotherapy replace surgery-first, and can pathological response replace long-term endpoints?
Target: PD-1
NADINA, CheckMate 816, NICHE-2 and KEYNOTE-522 show deep pathological responses, but regulators still want event-free or overall survival and surgeons want to know who can safely have less surgery.
PD-1
How long should checkpoint inhibitors be given: two years, one year, or until a ctDNA or imaging signal says stop?
Target: PD-1
Trials fixed two years by convention; long-term CheckMate 067 and KEYNOTE-006 data show durable remissions after stopping, and the cost and toxicity of unnecessary years are large.
PD-1
What actually turns an immune-desert tumour into one that responds to PD-1 blockade?
Target: PD-1
Most patients have primary resistance: no T-cell infiltrate or T cells excluded by stroma. Radiation, oncolytic viruses, ADCs and vaccines each prime in some models, but no combination has produced a reliable conversion in people.
PD-1
Can PD-L1 scoring be harmonised across antibodies (22C3, SP263, 28-8, SP142) and cut-points (TPS, CPS, IC) so one test serves every drug?
Target: PD-L1
Each drug was approved with its own assay and threshold; concordance is good for tumour-cell scoring but poor for immune-cell and combined scores, so patients are classified differently by which kit the lab bought.
PD-L1
Will personalised neoantigen vaccines confirm their phase 2 melanoma benefit in phase 3, and can manufacturing time and cost fall enough for routine use?
Technology: Personalised neoantigen (mRNA) vaccines
KEYNOTE-942 showed fewer recurrences with intismeran plus pembrolizumab; phase 3 (INTerpath-001) is the test, and six-to-eight-week manufacturing per patient limits use in fast-moving cancers.
Personalised neoantigen (mRNA) vaccines
Can degraders reach the transcription factors and scaffolds that inhibitors cannot, beyond the first approvals against oestrogen and androgen receptors?
Technology: PROTACs & molecular glues (targeted protein degradation)
Vepdegestrant showed degraders can match small-molecule antagonists; the promise is undruggable proteins (MYC, mutant p53, fusion transcription factors), where ligands for the target are the bottleneck.
PROTACs & molecular glues (targeted protein degradation)
For which adult cancers does proton therapy reduce toxicity or improve survival enough to justify its cost over modern photon radiotherapy?
Technology: Proton therapy
Dosimetric advantages are clear; randomised evidence in adults is limited to a few sites (oesophagus, head and neck, prostate under way), and payers differ widely in coverage.
Proton therapy
Should PSMA radioligand therapy move to hormone-sensitive disease, and is an alpha emitter better than lutetium-177?
Target: PSMA
VISION and PSMAfore established lutetium-177 PSMA-617 late in the disease; PSMAddition tests it in hormone-sensitive disease and actinium-225 programmes report deeper responses with salivary toxicity, but head-to-head data are absent.
PSMA
Should radioligand therapy be dosed by individual dosimetry rather than fixed activity?
Technology: Radioligand therapy (beta emitters)
Every approved radioligand is given as a fixed activity; dosimetry studies show several-fold variation in tumour and kidney absorbed dose, but no randomised trial has shown that personalised dosing improves outcomes.
Radioligand therapy (beta emitters)
Which targets can T-cell engagers hit in solid tumours without on-target toxicity in normal tissue?
Technology: T-cell engagers (bispecific)
Tarlatamab (DLL3) and tebentafusp (gp100) proved the principle; most solid-tumour antigens are shared with normal tissue and the therapeutic window depends on affinity tuning and masking.
T-cell engagers (bispecific)
Can actinium-225 supply scale to treat tens of thousands of patients a year if the phase 3 trials succeed?
Technology: Targeted alpha therapy
Actinium-225 comes from a handful of thorium-229 generators and accelerator routes with radionuclidic impurities; current output is a small fraction of projected demand.
Targeted alpha therapy
Does TIL therapy work outside melanoma, and can selection for neoantigen-reactive T cells raise response rates?
Technology: TIL therapy
Lifileucel is approved for melanoma after PD-1 failure; responses in lung and cervical cancer are lower, and most infused cells are not tumour-reactive.
TIL therapy
Can mutant p53 be drugged directly, whether by reactivation (Y220C binders), degradation, or synthetic lethality with WEE1 or ATR?
Target: TP53
TP53 is the most mutated cancer gene and still has no approved drug; eprenetapopt failed, Y220C reactivators are in trials, and WEE1 and ATR inhibitors have shown activity mainly in small studies.
TP53
Is there any biomarker (IHC, PET, quantitative continuous scoring) that predicts who benefits from a TROP2 ADC?
Target: TROP2
TROP2 ADCs are given without testing because IHC did not separate responders in ASCENT or TROPION-Breast01; TROPION-Lung01 reported a computational membrane-ratio score retrospectively but it is unvalidated.
TROP2

Framing sources are papers or registry records, not proof that the question is unanswered everywhere; if you know a trial or paper that settles one, suggest an edit. Questions appear on the target dossiers. Stage and actor use the same vocabulary as ideas, so a funder can filter both by who has to move.