The (Genomics) World Is Not Enough
Has precision medicine’s focus on the genome blinded us to vital non-genomic tools?
Joe Olechno |
I just finished reading The Pathologist’s “Tackling the Ticking Treatment Clock” (1), which I fully agree with. If a test is fast and inexpensive, why not use it? Slow tests delay treatment and allow both new mutations and metastasis. So I’m baffled that a procedure used at the Institute for Molecular Medicine, Finland (FIMM) has had so little impact. It can help save money and lives today.
Precision medicine (PM) is a step forward from the “standard-of-care” medicine of the past. Until recently, doctors used symptoms, bloodwork and experience to determine the best therapeutic approach. PM eliminates some of the variables and reduces the need for extensive experience. Yet genomics-based PM has run into some problems (2)(3)(4)(5). First, it is neither precise nor personalized. PM strives to link a DNA mutation to a successful treatment. It attempts to categorize the patient as a member of a specific group of patients. Within this group, patient therapy has shown positive results for the majority of its members. In other words, personalized medicine is actually population-based. As with other population-based groups, individual differences may overwhelm group similarities.
There are three critical problems with genomics-based precision medicine.
- DNA mutations are not clear signs of the cause of a particular cancer. PM uses genomic analysis to determine two things. First, which mutation causes a specific cancer, and second, the appropriate treatment based on the disease’s profile. Sometimes, the suggested treatment works. But other patients go through chemotherapy only to find that the cancer remains – or worse, they may find that it has grown and mutated during treatment so that it’s harder to treat. The promise of PM is that if you sequence the tumor and find a genetic flaw linked to the cancer, addressing that flaw should result in remission. Yet, despite the apparent causal relationship, there may be poor correlation between mutation identification and successful treatment. Researchers have shown that even healthy skin tissue is full of somatic DNA mutations (6) – as many as we see in tumor cells. There is also a strong positive selection for the mutations, so their numbers tend to increase. This means that a “cancer-causing mutation” may not actually cause a particular cancer – it may just be along for the ride. Targeting that mutation with chemotherapy will not only damage healthy cells, but delay the use of a drug that will actually work. Evidence of the existence of a mutation is not evidence of a cause for cancer.
- Reconciliation of genomic data with phenotypic results is difficult. The Cancer Genome Project (CGP) and the Cancer Cell Line Encyclopedia (CCLE) analyzed the DNA of hundreds of different cell lines and tested them against many different drugs to see how they responded. Both organizations had similar DNA results – but the results of the drug response tests were very different (7). The discordant results suggest that it may be difficult to assess gene–drug interactions. Just knowing the genetic variants a cancer cell carries is not enough to determine which drugs will work against it.
- When there is no drug for a particular mutation, personalized medicine reverts to “standard-of-care.” Not every cancer-causing mutation has an associated treatment. Without suggested treatments, the doctor bases therapy on past experience and clinical standards. There is no good way to determine what alternatives might be better. PM works only after researchers identify successful treatments.
What can we do instead?
Why not work backwards? Instead of identifying genetic mutations and prescribing drugs based on those mutations, researchers at FIMM (8–15) isolate cancer cells from each patient. They test these cells against hundreds of drugs to see which ones – or which combinations – work. Although not every drug that works in ex vivo functional screening will work in patients, those that don’t work in the screen are unlikely to work in the body. This knowledge allows the FIMM researchers to bypass drugs that wouldn’t help patients, sparing them the physical, psychological, and financial costs of failed chemotherapy while increasing the likelihood of a successful outcome. Rather than mis-labeled precision or personalized medicine, this is truly individualized medicine. It’s a clear application of medicine’s guiding rule: primum non nocere – first, do no harm.
While PM acknowledges non-genomic assays, its resources are largely devoted to genomic markers. That’s not to say that there is not a key role for genomic analyses in healthcare – there certainly is. Re-evaluating genomic mutations in the light of functional screening will provide new drugs and a better understanding of cancer. I just don’t believe that genetic analysis is the sole path to successful cancer treatment. I worry that, in focusing so intently on the genome, we may develop tunnel vision and ignore new tools – including new and modern ways of functional screening. This has the potential to impact millions of existing and future patients.
We need to stop turning a blind eye to non-genomic personalized medicine – and soon.
- M Schubert, The Pathologist, 10, 10 (2016).
- V Prasad, Nature, 537, S63 (2016). PMID: 27602743.
- MJ Joyner (2015). Available at: nyti.ms/1ByR2cA. Accessed December 5, 2016.
- SL Van Driest et al., JAMA, 315, 47–57 (2016). PMID: 26746457.
- L Husten (2016). Available at: bit.ly/1n4hhYP. Accessed December 5, 2016.
- I Martincorena et al., Science, 348, 880–886 (2015). PMID: 25999502.
- B Haibe-Kains et al., Nature, 504, 389-393 (2013). PMID: 24284626.
- T Pemovska et al., Cancer Discov, 3, 1416–1429 (2013). PMID: 24056683.
- J Tang et al., PLoS Comput Biol, 9, e1003226 (2013). PMID: 24068907.
- B Yadav et al., Sci Rep, 4, 5193 (2014). PMID: 24898935.
- V Pietiainen et al., Comb Chem High Throughput Screen, 17, 377–386 (2014). PMID: 24661208.
- PO Pietarinen et al., Blood Cancer J, 5, e309 (2015). PMID: 25933373.
- E Kulesskiy et al., J Lab Autom, 21, 27–36 (2016). PMID: 26721820.
- S Eldfors et al., Leukemia, [Epub ahead of print] (2016). PMID: 27461063.
- T Pemovska et al., Nature 519, 102-105 (2015). PMID: 25686603