Frequently Asked Questions

1. What is cancer biomarker testing?

Biomarker testing looks for features of a tumor that can help guide care, such as:

• Targeted therapy options (drugs aimed at specific tumor drivers)
• Immunotherapy guidance (in certain cancers)
• Clinical trial matches
• Diagnosis/subtype clarification (what kind of cancer this is)
• Prognostic or resistance information (how the tumor might behave or respond)

Biomarker testing may be done on tumor tissue (biopsy/surgery specimen) and/or blood (liquid biopsy).

2. What is NGS?

NGS (Next-Generation Sequencing) is a technology that reads genetic material at scale. In cancer testing, NGS commonly analyzes:

• DNA from tumor cells (or tumor DNA circulating in blood)
• RNA from tumor cells (which reflects what genes are being actively expressed)

NGS can examine a focused panel of genes or a broader profile, depending on the test.

3. What is the difference between DNA testing, RNA testing, and IHC?

These tests measure different biological layers of the tumor:

Key idea: DNA and RNA are about the genetic instructions and messages. IHC shows the protein output and where it appears in the tissue.

4. What does DNA NGS testing detect?

DNA NGS is especially good at identifying:

• Single nucleotide variants (SNVs) — single “letter” changes
• Small insertions/deletions (indels)
• Copy number alterations (amplifications or deletions of genes)
• Some rearrangements (depending on assay design)
• Genomic signatures such as MSI (microsatellite instability) and TMB (tumor mutational burden), when included

A typical DNA report includes a list of genetic alterations categorized as actionable, resistance, prognostic, or VUS (variant of uncertain significance), and sometimes MSI/TMB summaries.

5. What does RNA NGS testing detect?

RNA NGS is especially strong for changes best seen at the transcript level:

• Gene fusions — two genes abnormally joined together, creating a new driver
• Splice variants — abnormal editing of RNA that changes the resulting protein
• Confirmation of expressed events — helpful when DNA suggests a rearrangement but the exact breakpoint is hard to resolve

Why RNA can matter: Some clinically important drivers are fusions or splice events that may be missed or less clearly detected on DNA-only assays, depending on how the test is built and what regions are covered.

6. What does IHC testing detect?

IHC stains tumor tissue to show whether specific proteins are present (and where). IHC is commonly used for:

Therapy-related protein biomarkers

• PD-L1 (often used to help guide immunotherapy decisions)
• HER2 (protein overexpression)
• ER/PR (commonly in breast and some gynecologic cancers)

Mismatch repair protein testing (MMR)

IHC can measure MLH1, MSH2, MSH6, PMS2 protein expression. Loss of one or more can suggest dMMR, which often correlates with MSI-high biology.

Tumor identification and classification

IHC is a core pathology tool for determining tumor lineage, likely site of origin, and subtype refinement that affects treatment pathways.

Why IHC is unique: It provides spatial context — it can show whether a protein is present specifically in tumor cells, immune cells, or surrounding tissue.

7. Why might it be important to have both DNA and RNA testing?

Because each can find important drivers the other might miss:

• DNA excels at: mutations, copy number changes, and DNA-derived signatures (often MSI/TMB if included)
• RNA excels at: fusions and splice variants, and proving an event is expressed

Practical examples: DNA may show no clear driver mutation, but RNA finds an actionable fusion. Or DNA may detect a rearrangement signal, while RNA clarifies the exact fusion transcript.

8. Why add IHC on top of DNA/RNA?

IHC often answers questions neither DNA nor RNA can fully answer on their own:

• Is the relevant protein actually present (or absent) in tumor cells?
• How strong is expression, and in what fraction of cells?
• Is this tumor type what we think it is? (critical for choosing the right treatment pathway)

In many real-world workflows, IHC and NGS are complementary rather than redundant.

9. Tissue biopsy vs liquid biopsy

Tissue-based testing

• Uses tumor from a biopsy or surgery
• Can support DNA NGS, RNA NGS, and IHC
• Often provides the most complete picture, especially when tissue is adequate

Liquid biopsy (blood-based)

• Usually tests circulating tumor DNA (ctDNA) in the blood
• Convenient when tissue is hard to obtain or when rapid assessment is needed
• Important limitation: if the tumor is not shedding much DNA into blood, results can be negative even when a tumor driver exists

10. If a test is “negative,” does that mean there are no actionable biomarkers?

Not necessarily. A negative result can happen for several reasons:

• Low tumor content in the sample
• Degraded material (especially RNA in older or poorly preserved tissue)
• Tumor heterogeneity (different tumor sites can have different drivers)
• Low ctDNA shedding in blood-based tests
• The driver exists but is outside the test's coverage or is a complex alteration

This is one reason clinicians sometimes use reflex testing (e.g., add RNA fusion testing after a DNA-only test, or pursue tissue testing after a negative liquid biopsy).

11. What do common result categories mean?

• Actionable / therapeutic: may support an approved therapy or a strong guideline-backed option
• Clinical trial: may qualify for a trial targeting that alteration
• Resistance: suggests lower likelihood of response to a therapy
• Prognostic: associated with outcomes but not necessarily treatment-selecting
• VUS (Variant of Uncertain Significance): a change detected, but not currently known to affect treatment

12. Somatic vs germline: why does it matter?

• Somatic alterations are in the tumor only.
• Germline alterations are inherited and present in normal cells too.

Some tests compare tumor DNA to a matched normal sample (like blood/saliva) to help separate somatic from germline and reduce false positives.

13. How do testing companies differ?

Different testing companies and products (Foundation Medicine, Caris, Tempus, Guardant, etc.) can vary in:

• Specimen type: tissue vs blood
• Modalities included: DNA only vs DNA+RNA; whether IHC is included or ordered separately
• Breadth: targeted gene panel vs broader profiling
• Fusion strategy: DNA-based fusion calling vs RNA-based fusion detection
• Reporting style: actionability tiers, trial matching, evidence summaries

Rule of thumb: don't assume two NGS tests are equivalent — ask what's actually included (DNA? RNA? IHC? MSI/TMB? fusion/splice coverage?).

14. When might you specifically ask about RNA fusion/splice testing?

RNA testing is often especially relevant when:

• The cancer type commonly has actionable fusions (many lung cancers, certain sarcomas, cholangiocarcinoma, thyroid cancers, and others)
• DNA testing was negative or inconclusive but suspicion remains for a driver
• A rearrangement is suspected and needs clearer resolution
• A splice event is clinically important for the tumor type

15. Questions to ask your oncology team or testing lab

Use these to confirm you are getting complete coverage:

1. Was the test DNA-only, RNA-only, or both?
2. Does it include RNA-based fusion and splice detection?
3. Were MSI and TMB reported (if relevant for my cancer type)?
4. Was IHC done for key markers (PD-L1, HER2, ER/PR, MMR), if relevant?
5. Was the sample adequate (tumor % / quality metrics)?
6. If negative, should we reflex to RNA, repeat on another tissue site, or consider liquid biopsy (or vice versa)?
7. Are any findings VUS, and do they change anything today?
8. Do the results open clinical trial options?

16. Glossary

• NGS: Next-generation sequencing
• DNA: genetic blueprint (mutations, amplifications, deletions)
• RNA: gene transcripts (fusions, splice variants, expression)
• IHC: protein staining in tissue (expression + spatial context)
• Fusion: abnormal joining of two genes that can drive cancer growth
• Splice variant: altered RNA processing that changes the resulting protein
• MSI/dMMR: DNA repair deficiency that can influence immunotherapy response (context-dependent)
• TMB: tumor mutational burden; sometimes used in immunotherapy decision-making (context-dependent)
• ctDNA: circulating tumor DNA found in blood
• VUS: variant of uncertain significance

1. What is ctDNA?

Tumor cells shed fragments of their DNA into the bloodstream. These fragments are called circulating tumor DNA (ctDNA). A simple blood draw can capture and analyze these fragments without requiring a tissue biopsy.

ctDNA carries the same genetic mutations as the tumor itself, which means it can be used to detect cancer-specific alterations, monitor how much tumor is present, and track changes over time.

2. ctDNA during active treatment (therapy monitoring)

Serial ctDNA blood tests taken during treatment can show whether tumor burden is rising or falling — sometimes weeks before imaging changes are visible.

• Rapid drop in ctDNA after starting therapy may indicate early response
• Persistent or rising ctDNA during treatment may signal resistance or progression
• New mutations appearing in ctDNA can reveal emerging resistance mechanisms

This real-time molecular feedback can help oncologists assess whether a treatment is working without waiting for the next scheduled scan.

3. ctDNA after treatment — MRD detection

For patients who are NED (no evidence of disease) after surgery, radiation, or systemic therapy, ctDNA testing can detect molecular residual disease (MRD) — cancer that may still be present at levels too small for CT scans, MRIs, or PET scans to find.

• MRD-positive: Tumor DNA is still detectable in blood even though scans are clean. This can identify patients at higher risk of recurrence.
• MRD-negative: No tumor DNA detected. This is encouraging but does not guarantee the cancer will not return — some cancers shed very little ctDNA, and the test has a detection threshold.

4. How MRD results are being used

In some cancer types, MRD testing is being studied to guide treatment decisions such as:

• Whether to add adjuvant chemotherapy after surgery (for MRD-positive patients)
• Whether it is safe to de-escalate or shorten treatment (for MRD-negative patients)
• When to intensify surveillance or consider additional therapy
• Whether to enroll in a clinical trial designed around MRD status

Cancer types with active MRD research: colorectal, lung (NSCLC), breast, bladder, esophageal, pancreatic, and others. This is one of the fastest-moving areas in oncology — ask your oncologist whether MRD-guided trials are available for your cancer type.

5. Available ctDNA/MRD tests

Several commercial tests are available, and they differ in important ways:

Tumor-informed tests first sequence your tumor tissue to identify your specific mutations, then build a personalized blood test to track those exact mutations. Tumor-naive tests look for common cancer signals without needing prior tissue sequencing. Tumor-informed tests are generally considered more sensitive but require available tumor tissue.

6. Important limitations

• Not all cancers shed ctDNA reliably — brain tumors (due to the blood-brain barrier), some low-grade cancers, and mucinous tumors may produce little detectable ctDNA
• A negative result does not prove absence of disease — the cancer may be present below the test's detection threshold
• MRD-guided treatment decisions are still being validated in clinical trials for many cancer types — standard-of-care guidelines have not yet incorporated MRD results universally
• Insurance coverage is inconsistent — some insurers cover ctDNA monitoring for specific cancer types and stages, while others consider it investigational
• Clonal hematopoiesis (CHIP) — age-related blood cell mutations can sometimes be confused with tumor-derived ctDNA, though newer tests are designed to filter these out

7. Questions to ask your oncologist

1. Is ctDNA monitoring relevant for my cancer type and stage?
2. Should I have MRD testing after completing treatment?
3. How would a positive or negative MRD result change my treatment plan?
4. Are there MRD-guided clinical trials I might qualify for?
5. Which ctDNA test would be most appropriate — tumor-informed or tumor-naive?
6. How often should ctDNA be tested, and for how long?
7. Will my insurance cover this test?

8. Glossary

• ctDNA: circulating tumor DNA — tumor DNA fragments found in blood
• MRD: molecular residual disease — cancer detectable by molecular testing but not by imaging
• NED: no evidence of disease — no cancer visible on scans
• Tumor-informed: MRD test customized using your tumor's known mutations
• Tumor-naive: MRD test that does not require prior tumor sequencing
• CHIP: clonal hematopoiesis of indeterminate potential — age-related blood cell mutations unrelated to cancer
• Adjuvant therapy: treatment given after primary treatment (e.g., surgery) to reduce recurrence risk

Data sources

• CIViC (Clinical Interpretation of Variants in Cancer) — an open, expert-moderated knowledgebase of clinical variant interpretations drawn from published literature. Its nightly evidence export is our discovery source for new drug–biomarker–disease relationships; we consume evidence at CIViC's two highest tiers (Level A, validated; Level B, clinical).
• openFDA — FDA structured product labeling (indications and usage) and Drugs@FDA marketing status. This is our source of truth for every approval claim: evidence can propose, but only the label and an active marketed product can confirm.
• ClinicalTrials.gov via the AACT database, refreshed daily, for trial matching.
• PubMed, refreshed nightly, for the published-research features.

The weekly evidence cycle

Every Monday, new and changed CIViC evidence is diffed against the prior week and routed down one of three paths:

• Automatic additionis reserved for the strictest case: validated (Level A) evidence, for a biomarker the platform already recognizes, where the drug's current FDA label and an active marketing status are both independently confirmed via openFDA at the moment of addition. Every row then passes a single shared validator that rejects contradictory states (for example, an approval flag on an off-label indication, or a combination regimen matchable by a single component).
• Existing drug, proposed new cancer type — the most error-prone category in practice — is never applied from evidence alone. Each proposal is individually adjudicated against the FDA label (see the review layer below), because published evidence of activity in a disease is not the same thing as an approved indication for it.
• Everything else goes to review rather than to the catalog.

Approval status is modeled per indication, not per drug: a therapy can be on-label for one cancer type, off-label for another, and explicitly excluded for a third (label Limitations of Use are representable and displayed), with clinical notes carried on the specific indication.

The AI review layer, deliberately constrained

Candidates that are neither automatic accepts nor automatic rejects are adjudicated by a large-language-model review step that functions like a junior curator with a strict protocol. It is given the drug's actual FDA label text (retrieved live from openFDA) together with grounded literature search, and must return a verdict — approve, reject, or uncertain — with a written rationale and citations. Two independent evaluation passes must agree, a conservative confidence bar applies, and anything below it, or marked uncertain, routes to a human queue rather than to the catalog. Verdicts are remembered with their provenance; negative and uncertain calls are periodically re-examined, because approval status changes; and a human override is permanent — the model never re-litigates it.

Continuous re-verification

• Weekly (Sunday): the entire catalog — not just new entries — is re-checked against openFDA. Market withdrawals and discontinuations are detected this way (a withdrawn drug is deactivated with a visible explanation, never silently deleted), and evidence-only entries that have since gained approval are flagged for upgrade. This sweep is read-only by design; resulting changes ship only through reviewed code.
• Daily: an operational health report verifies that scheduled runs actually happened and that both CIViC and openFDA are reachable, with live probes. A missing pipeline run is itself treated as an alarm — silence is a failure signal, not a pass.

Vocabulary control

Cancer types and biomarkers resolve against curated canonical vocabularies. Incoming disease strings normalize through an alias layer (so "appendix cancer," "appendiceal adenocarcinoma," and "pseudomyxoma peritonei" land on one canonical type); strings that don't resolve are proposed for human mapping, never auto-created. Histologic subtype mappings are exact-match allow-lists with an asymmetric safety model: an omission can only reduce precision, never produce a wrong match. Gene symbols are normalized to canonical forms so that notation differences (PML::RARA versus PML-RARA) cannot split a patient from a matching therapy.

Human oversight, on the record

Every AI verdict — including rejections — is reported weekly with its rationale and citations, in a decision-first format that separates items awaiting human judgment from items already resolved. Removals always require human confirmation. Every catalog row carries provenance (what added it, from which evidence, when), changes are reversible, and the same validation applies on every write path, human or automated.

Why this instead of a manual curation team

It is a fair question. The answer is that the upstream expert judgment already exists and is continuously maintained by the clinical community: CIViC's interpretations are expert-submitted and moderator-reviewed against published literature, and FDA labeling is the regulatory ground truth. What a paid curation team would add on top of those sources is surveillance, cross-verification, consistency, and record-keeping — which are precisely the tasks software does better than people: this pipeline re-checks every entry weekly, verifies every claim against the current label at the moment it is made, never gets tired on item 400 of a diff, and leaves a complete audit trail. The genuinely human decisions — ambiguity, clinical nuance, removals, vocabulary — are exactly the items the system refuses to decide and routes to review. And structurally, the process is manufacturer-blind: listings are driven by public evidence and FDA labeling applied identically to every company's products.

None of this replaces clinical judgment, and it is not presented as such: everything a patient sees is framed as a starting point to review with their treating oncologist.