The Precision of Computational Medicine

In a recent Question & Answer session, Ryan Rossier, VP of Product Innovation at Medullan, a digital healthcare consultancy based on Cambridge, Mass., discussed the impact of health information technology on personalized medicine.

Q: Can information technology, big data, and data mining make precision medicine a reality?

Rossier: Yes! It’s already happening, through the dedication of large and small technology companies alike. We can do this today due to the confluence of factors, including the persistence of Moore’s Law and its impact on computing power, storage capacity, and bandwidth throughput. There’s a tremendous amount of data coming online through the required use of electronic medical records, increased health system interoperability, the advancement and capability of sensors in consumer-grade connected devices and smartphones, and the cost-decay we’re seeing with personal genome sequencing. Add the ability to develop artificial intelligence and machine-learning software (data mining) and you’ve got “Computational Medicine” which aims to personalize treatments and create tools for greater precision and efficacy in healthcare.

Some examples would include:

  • oncology for personalizing treatments – Imagine an oncologist reviewing a case for a patient that has tested positive for colon cancer. She wants to prescribe a common chemotherapy, Irinotecan, to her patient. In the case that this drug is prescribed without a thorough understanding of the patient, potentially severe and life threatening side effects could occur. The oncologist first wants to check to see if her patient has a variant that would cause them to metabolize this drug more slowly than others. If this patient has had their genome sequenced already, the oncologist simply has to look it up… it could be as easy as checking their PHR. If not, she could run a genetic screening specifically for this variant. Alternatively, she could tap into a database of people similar to this patient (by gender, height, weight, ethnicity, region, etc.) and make an educated dosage decision based on trends; and
  • radiology for increasing efficiency: – Images are such a large part of medicine today. At the doctor’s office you will hear “I’ll have Doctor Smith look over your X-ray and come in”. When a radiologist looks over the image they’re using all of their past experience, education, and training to evaluate the pattern they’re seeing in the image, which can often be subtle. But, this is a perfect example where software and particularly image recognition and machine learning when paired together can do better.

Consider a computer algorithm that does the equivalent of what spellcheck does in word processing, but for these medical images. In that case, the software reviews thousands of words, but only highlights the ones which are worth reviewing, based on its rules and learned pattern recognition. The solution doesn’t intend to replace every radiologist, but it aims to drastically increase the efficiency of the workflow, allowing HCPs to focus their energy on more important and complex cases.

Q: What’s the best way to proceed?

Rossier: The goal is to put data and insights directly in the hands of Healthcare Professionals. Give them tools that are easy to use, fit into their clinical workflow, and specifically target their most common questions. e.g., “What’s the most effective medicine/dosage to prescribe for ‘XYZ’ condition?” “Will the person I’m training react better to endurance exercises or strength training exercises?” and “Does this patient have an allergy to a particular food or show a sensitivity I need to consider?”

  • Educate physicians on the value of sequencing and how to communicate with patients about their genome
  • Motivate patients to opt-in, get their genomes sequenced, and get involved in research. Precision medicine solutions feed off of data, and will get smarter and more powerful the more data they consume, so let’s give them everything they can handle.
  • Ultimately, treat genome sequencing as a routine lab test. I go to my annual exam, my PCP checks my lab history and notes that it’s time for a lipid panel, and also notes that I haven’t had my full genome sequenced yet and adds that too. I go to the facility, spit in a cup, and my results are all sent to me and my PCP. I’m notified weeks later of my combined results and now my genome is codified for future use.
  • And an initiative gaining momentum with my colleagues at Medullan: get all babies sequenced at birth, with an opt-out policy.

Q: Will it improve clinical outcomes?

Rossier: As far as we can hypothesize, computational medicine has the potential to drastically improve clinical outcomes. I believe it also has the opportunity to also improve the patient experience at the same time.

For outcomes, consider the deaths that occur from misdiagnosed conditions. Consider the dangerous side effects that result from high dosages of a medication I can’t tolerate. Consider the ability to know at birth if my baby has a life threatening condition that we can treat BEFORE signs appear.

For the patient experience: There’s nothing more specific to patient than their own DNA-makeup, so I believe leveraging an individual’s genome as a key input to determining treatment plans has the potential to connect patients more closely to their overall healthcare experience, their condition and the resulting treatment plans, especially if combined with an overall experience that’s easy to understand and well supported.

Content brought to you by ADVANCE and Medullan.

About The Author