Today, NGS (Next-Generation Sequencing) tests that map a tumor’s genetic landscape have become standard in advanced cancer. FoundationOne CDx, Guardant360, Tempus, Caris and other panels generate reports dozens of pages long. But the real clinical challenge isn’t running the test: it’s reading the report correctly, recognizing its gaps and pitfalls, and translating it into a patient-specific treatment roadmap.
An NGS report tells you which mutations are present. It does not tell you which treatment is appropriate or inappropriate for this patient, in this sequence, and why. This is one of modern oncology’s biggest problems: we have more and more data, yet turning that data into a patient-specific decision is left largely to the treating physician.
A report usually offers a list of variants and occasional drug associations, but it doesn’t prioritize them; it doesn’t separate which finding is a true driver from which is a weak or misleading signal; it doesn’t build context by tumor type and line of therapy; and it often doesn’t say what not to do. On top of that, while approvals and evidence shift within months, the physician is expected to correctly interpret dozens of pages of data in limited time. The result: the same report can lead to an ineffective or inappropriate treatment when misread, or to a life-changing opportunity when read correctly.
A molecular second opinion exists precisely to close this gap: to build the bridge between raw genomic data and an actionable, reasoned, patient-specific treatment roadmap.
Prof. Dr. İbrahim YıldızSummary: the cases at a glance
Below, I share thirteen cases we consulted on (with patient identity fully anonymized). Each one matters because it diverges from the standard report at a different decision point.
| # | Tumor / Context | What the second opinion changed |
|---|---|---|
| 1 | Intrahepatic cholangiocarcinoma | DNA looked "clean" while an FGFR2 fusion was caught only in RNA → a missed targeted therapy |
| 2 | Hepatobiliary adeno (BRAF) | Distinguishing BRAF Class III (G464E) → avoiding the wrong-BRAF-inhibitor trap |
| 3 | Low-grade serous ovarian | Endocrine therapy by hormone-receptor positivity instead of an unnecessary toxic/inaccessible combination; "what not to do" clarity |
| 4 | NSCLC (KRAS G12R) | Identifying the best immunotherapy subgroup in a case thought "untargetable" |
| 5 | Urothelial / bladder (liquid) | Three traps in one report: subclonal CHEK2 → PARP rejected, TMB synthesis, FGFR1 amp ≠ erdafitinib |
| 6 | Pancreas (KRAS G12D, 3rd line) | Prioritizing daraxonrasib based on recent positive phase 3 data |
| 7 | SMARCA4 (DNA/IHC conflict) | Absent in DNA, lost on IHC → epigenetic mechanism; methylation test recommended |
| 8 | Prostate (liquid) | Putting unprioritized options in the right order (PTEN/AKT first, not zenocutuzumab) |
| 9 | Bladder urothelial (pretreated) | High PD-L1 → immunotherapy priority; PARP rejected on weak CHEK2; contextualizing off-label MET |
| 10 | HER2 (RNA/DNA conflict) | Ordering the right test (IHC/FISH) for anti-HER2 |
| 11 | Leiomyosarcoma | Translating an RUO RNA expression signature into chemotherapy selection |
| 12 | RET-fusion lung adeno | Anticipating SCLC transformation and bypass resistance before treatment starts |
| 13 | Colorectal | Correctly interpreting a conflicting MSI-equivocal call against low TMB |
The DNA panel said "clean," yet the most critical target was being missed
No targetable mutation was found on the tumor’s DNA panel; the report looked practically "empty." Had that result been trusted alone, the patient would have been planned for standard chemotherapy only.
But RNA-based fusion analysis of the same sample detected an FGFR2 fusion, the direct target of FDA-approved targeted drugs (FGFR inhibitors) in bile duct cancer.
Why didn’t it show in DNA? A significant share of FGFR2 fusions cannot be captured by a DNA panel because the breakpoint falls in a region the panel probes don’t cover; RNA, by reading the mature transcript, shows the fusion directly.
Interpreting the multi-platform test (DNA + RNA) together opened a life-extending targeted therapy option that would otherwise have been missed entirely. Our recommendation is clear: in bile duct cancers DNA alone is insufficient: RNA fusion analysis must be considered mandatory.
Same gene, very different meaning: the BRAF trap
The report said "BRAF mutation." The first reflex might be to consider the BRAF inhibitors familiar from melanoma/colorectal cancer. But this mutation was not the classic BRAF V600E. It was a Class III (non-V600) BRAF mutation (BRAF p.G464E), a subtype with low kinase activity that signals through RAS-dependent dimers (CRAF).
As a result, classic (Class I) BRAF inhibitors designed to target V600E monomers (vemurafenib, dabrafenib, encorafenib) are ineffective alone against this mutation and can paradoxically activate the MAPK pathway.
We emphasized that classic (Class I) BRAF inhibitors should be avoided in this profile; the rational target is the MEK-inhibition axis. We also flagged relevant early-phase trial avenues for the accompanying MTAP deletion and ARID1A/KDM6A losses.
The subtype distinction behind a single gene name prevented a misdirected targeted therapy.
Defining not just "what to do" but clearly "what not to do"
Low-grade serous ovarian cancer is relatively chemo-resistant; the expected benefit from a classic cytotoxic approach is limited. The report showed a dominant KRAS mutation. Standard drug-mutation matching would highlight a MEK/RAF-axis targeted combination based on the KRAS driver, but that combination was neither accessible for this patient nor anything but unnecessary and toxic.
Because of ER and PR expression (hormone-receptor positivity) detected in the RNA layer of the NGS, we recommended far better-tolerated hormonal suppression (endocrine therapy); low-grade serous ovarian cancer being frequently hormone-sensitive supports this choice. A single-agent MEK inhibitor was also noted as an option.
Immunotherapy was not recommended due to low TMB, MSS, and PD-L1 negativity.
Short rationale: in a setting where chemotherapy doesn’t work and a toxic/inaccessible targeted combination is unnecessary, the tumor’s own biology (hormone-receptor positivity) pointed to the least toxic and most feasible path.
Correctly positioning a mutation thought to be "untargetable"
The patient had a KRAS G12R mutation, a variant with no allele-specific approved drug as yet. At first glance it can look "untargetable."
KRAS was accompanied by TP53, with no STK11/KEAP1, a combination that defines the molecular subgroup with the best immunotherapy response in KRAS-mutant lung cancer. We recommended a chemo-immunotherapy combination first-line as the priority, asked for PD-L1 testing to be completed urgently, and flagged a next-generation pan-RAS inhibitor trial for progression.
Molecular interpretation turned "no target" into the insight that "the strongest opportunity is actually immunotherapy."
One report, three different traps
Let’s state it upfront: the standard report (Guardant) recommended a PARP inhibitor (olaparib, talazoparib) for this patient based on a CHEK2 alteration. We rejected that recommendation. This case shows how we diverged at three separate decision points within a single report:
The VAF of the CHEK2 alteration was only 0.6%, far below the tumor fraction (16.2%), i.e., a deeply subclonal signal; the same patient had CHEK2-related clonal hematopoiesis (CHIP) variants that confounded the signal; and there is no CHEK2-based PARP-i approval in urothelial cancer. Result: an unnecessary and unsupported targeted therapy was avoided.
The standard report left a high TMB value (28.47 mut/Mb) without an associated therapy. We synthesized it together with ARID1A loss + TERT promoter mutation + urothelial histology into a coherent immunotherapy-favoring biology and converted it into a numeric IO-eligibility score, adding the caveat that "blood-TMB is not a tissue-validated analyte; tissue confirmation is recommended."
A clinician seeing the FGFR1 amplification in the report might reflexively consider erdafitinib. We warned explicitly: erdafitinib is approved only for FGFR3 mutation or FGFR2/3 fusion; FGFR1 amplification does not qualify (and here the amplification was low-level, copy number ~2.3). Avoiding giving the right drug to the wrong target, the highest-value type of contribution for safety.
The latest evidence: yesterday’s standard may not be today’s best
The standard NGS report recommended an older/general MEK-axis combination targeting KRAS (avutometinib + defactinib).
Based on very recently reported positive phase 3 data for this variant, we prioritized the pan-RAS inhibitor daraxonrasib, which directly targets RAS. In previously treated metastatic pancreatic cancer, this agent showed a significant overall-survival advantage over standard chemotherapy (RASolute 302 phase 3, KRAS G12-mutant population). The drug is still in the approval process and is accessible through an appropriate clinical trial.
The principle matters: in precision oncology evidence changes within months; a recommendation must align with the most current phase 3 literature at the moment the report is written. Standard vendor reports often don’t update at this pace.
Solving the mechanism where DNA and pathology conflict
There was a contradiction here: DNA sequencing showed no mutation/loss in SMARCA4, while immunohistochemistry (IHC) showed SMARCA4 protein loss.
Rather than glossing over the contradiction, we explained its mechanism: a protein disappearing without a detectable DNA change usually points to epigenetic silencing (promoter methylation) or other mechanisms a panel can’t capture. To resolve the contradiction we recommended an additional methylation test. The correct treatment decision can only be made after this confirmation.
This is a textbook example of not falling into the "if it’s not in the report, it doesn’t exist" fallacy.
Prioritizing the right one when there are multiple findings
The report had several alterations at once: AR amplification, PTEN biallelic loss, an FGL1-NRG1 fusion, and a borderline-high blood-TMB. The standard report recommended capivasertib and zenocutuzumab side by side without prioritization; to the physician the two agents looked equally weighted.
PTEN biallelic loss → capivasertib (moved up): a genuinely actionable target with phase 3-level evidence in prostate cancer (AKT pathway inhibition) that directly matches the tumor type.
FGL1-NRG1 fusion → zenocutuzumab (not prioritized): its approved indications are NRG1-fusion NSCLC, pancreatic and bile duct cancers; there is no approval/efficacy data in prostate. Reliable confirmation also requires tissue-based RNA; this case was a liquid biopsy.
AR amplification was more a sign of resistance than a targetable opportunity. The blood-TMB was borderline and liquid-based; a conditional signal requiring tissue confirmation before any standalone immunotherapy decision.
Here the difference wasn’t finding a new target; it was doing the prioritization the standard report didn’t. Not promoting an attractive-looking but inappropriate option is also part of that guidance.
Setting the right priority in a heavily pretreated patient
The standard report pointed at a PARP inhibitor (high-VAF CHEK2), an inhibitor for MET amplification, and immunotherapy for high PD-L1 (22C3, CPS 60) all together, but didn’t prioritize.
Given this profile and treatment history, we prioritized immunotherapy (pembrolizumab) because of the high PD-L1. High PD-L1 with accompanying ARID1A loss was a strong checkpoint-sensitivity signal; it was the most robust, evidence-based option at this line.
A high VAF doesn’t make a variant a PARP target. CHEK2 p.I157T was a low-penetrance, founder-type missense variant and, at a ~66% purity-adjusted VAF, sat in the heterozygous germline range, not biallelic loss. A PARP inhibitor requires true HRD. We placed the high VAF in the context of germline confirmation and genetic counseling and referred to clinical genetics.
MET inhibitors (capmatinib/tepotinib) are approved in NSCLC but off-label in urothelial, to be considered only within an appropriate clinical trial, as a later-line option. Also, since FGFR3 was wild-type, erdafitinib didn’t apply; given RB1 + TP53 dual loss we recommended cautious surveillance for neuroendocrine transformation.
RNA says "present," DNA says "absent": ordering the right test for HER2
The same report held a contradiction: RNA sequencing showed HER2 (ERBB2) and ERBB3 overexpression, while the DNA panel showed no HER2 amplification.
The actionable and approved threshold for anti-HER2 therapy (e.g., ADCs such as trastuzumab deruxtecan) is not RNA overexpression but protein level, i.e., IHC, supported by FISH where needed. RNA overexpression alone is not a validated biomarker; the only correct way to resolve the contradiction was to order HER2 IHC/FISH.
If IHC 3+ / FISH+: an effective treatment door opens: trastuzumab deruxtecan (tumor-agnostic) or zanidatamab, which might otherwise have gone unnoticed. If not: an unnecessary and ineffective anti-HER2 therapy based on the RNA signal alone is avoided.
Cross-platform discordance (RNA vs DNA vs protein) is not "confusion"; managed correctly, it is a decision opportunity.
Translating an RNA expression signature into chemotherapy selection
The vendor’s report listed certain RNA-Seq expression values (TOP2A and CCNE1 high, MGMT low) only as Research-Use-Only (RUO) data and made no treatment recommendation from them. For the clinician these lines might have remained "numbers that don’t tell me what to do."
TOP2A high: the direct target of anthracyclines (doxorubicin); this molecularly supported the first-line doxorubicin-based regimen that is already NCCN Category 1.
MGMT low: a hypothesis of sensitivity to alkylating agents (dacarbazine/temozolomide); can be considered at a later line.
CCNE1 high: Cyclin E1 / CDK2 axis; rationale for an appropriate CDK2 inhibitor trial.
These RNA expression data are RUO; there is no validated companion diagnostic. We did not present them as "proven predictive biomarkers"; we labeled them clearly as hypothesis-generating biological rationale that prioritizes treatment.
The diagnostic value of RNA fusion analysis was also used: the absence of fusions such as SS18-SSX, EWSR1, NTRK supported the classification of fusion-negative high-grade leiomyosarcoma.
Anticipating resistance and histologic transformation before treatment begins
First-line treatment is clear: a RET-specific inhibitor (selpercatinib), NCCN Category 1. But the real contribution was a warning that anticipated what would happen later, before treatment even began.
Histologic transformation (SCLC) surveillance: biallelic TP53 inactivation and RET-driven instability create the ground for phenotypic transformation to SCLC. On progression, re-biopsy in case of unexpectedly rapid progression/neuroendocrine features; if transformation is confirmed, carboplatin + etoposide ± atezolizumab (IMpower133).
Bypass-pathway warning: ERBB3 (HER3) and MET overexpression on RNA-Seq pointed to bypass resistance via PI3K-AKT; an appropriate next-generation agent/trial plan was defined in advance.
Even though PD-L1 looked technically eligible, with a proven RET fusion single-agent immunotherapy is not recommended first-line (RET-fusion NSCLC is an "immune-cold" phenotype); the RET inhibitor takes priority.
The difference here is proactive: mapping out for the physician, while treatment is just starting, the form resistance might take.
Correctly interpreting the presence of a conflicting biomarker
The report returned an MSI status of "borderline/equivocal." A clinician might read this as a possible MSI-H signal, i.e., an immunotherapy opportunity.
We evaluated this signal not in isolation but together with TMB. The TMB was low (5.3 mut/Mb); dMMR/MSI-H strongly co-occurs with high TMB, so against a low TMB the most likely meaning of an equivocal MSI call is MSS/pMMR, and it must be re-confirmed first. We ordered MMR-IHC anyway. The accompanying MUTYH being in the germline range supported a non-hypermutated profile and was flagged for germline counseling.
The difference was checking biological consistency rather than jumping at a single "promising" line, and protecting the patient from being steered into an immunotherapy that probably wouldn’t work.
So why a molecular second opinion? What’s our difference?
These thirteen cases show one thing: an NGS report is a beginning, not a conclusion, and the contribution emerges at a different point in each case.
How do we make things easier for the physician and the patient?
For the treating physician: it reduces dozens of pages of raw data into a step-by-step (1st / 2nd / 3rd line) treatment matrix, a resistance-management plan, and a prioritized action list. Instead of hours of literature review, the decision is supported by a structured summary.
For the patient and family: it provides a clear, reasoned answer to "What does this test mean for me, what are my options, and why in this order?" Uncertainty decreases and decisions are shared.
Closing
In modern oncology the difference often lies not in running the test, but in reading the report correctly. In each of the cases above, the correct interpretation either rescued an opportunity that would have been missed, prevented an unnecessary/wrong treatment, opened a new path where standard treatment was exhausted, or refreshed the recommendation with the latest evidence.
If you have an NGS report in hand and the question "are we planning the right treatment, in the right sequence?" is on your mind, a molecular second opinion exists for exactly that.