Sequencing the first human genome was a massive public undertaking costing $2.7 billion and taking nearly 15 years to complete. The current cost of sequencing a human genome has dropped dramatically. The cost dropped from $4000 in 2015, to less than $300 today.
Some genomic companies are further aiming to drop the cost to less than $100. At these prices the question then shifts to when is it neglectful not to sequence a human genome?
One study found that early diagnosis of cancer could save the American medical system an average of $26 billion per year. In comparison sequencing every American would incur a one-time cost of $100 billion, and this cost could further drop if economies of scale were factored in.
Outside of finding genetic predispositions to cancer, genome sequencing is directly relevant to identifying diseases associated with multiple types of inherited genetic disorders such as single gene inheritance, multifactorial inheritance, chromosome abnormalities, and mitochondrial inheritance.
Now, let’s imagine the benefits to society if a subset of machine learning called deep learning was used to analyze this treasure trove of genetic information.
Loosely based on the brain’s neural networks, deep learning is one of the most important tools that data scientists rely on. Deep learning analyzes data and uses pattern recognition to create sophisticated models that can learn complex patterns and generalize those patterns to identify datapoints. Those datapoints can be complex and the more data a deep learning system operates with the more impressive the results.
By feeding an entire human population into a deep learning system, machine learning algorithms could identify potential biomarkers for cancer or other illnesses. Furthermore, the system could dive deep into genetic family trees to identify what genes are responsible for different traits or ailments.
None of this is revolutionary to practitioners of AI, the problem is our reliance on antiquated governments that rely on decades old technology, often times fax machines are still used to send patient data back and forth. It’s time to modernize this archaic system.
Now imagine if next time you went to your doctor, they could instantly access your genome, and feed your latest symptoms into a computer to instantly have the AI return results with recommendations for treatment options. With wearable devices that track real-time health data visiting a doctor may not even be necessary.
The most effective solution would be for healthcare staff to be trained in genetic counseling, this is the process of investigating individuals and their bloodlines who are affected by or at risk of genetic disorders. The most important step in genetic counseling is recommending the best treatment options based on that individuals genetic profile.
Another benefit is targeting neglected rare diseases, a rare disease is identified as a health condition that affects a small number of people compared to more common diseases that affect the general population. These rare diseases are often too costly to target the traditional way.
A deep learning system could compare datasets (genetic profiles) to potentially identify genetic precursors to these rare diseases. Rare diseases alone affect 30 million Americans and cost the US medical system $1 trillion a year.
One valid argument against the above would be the potential loss of user privacy. These concerns could be remedied by using an advanced type of machine learning called federated learning. With federated learning there is no need to share any personally identifiable data. Federated learning brings machine learning models to the data source, enabling data to be stored in a secured location. To reduce privacy concerns a simple solution would be for a patient (or parent) to approve access to personally identifiable data.
This type of solution could be vital for assisting with bringing personalized drugs to market. Knowing what molecules to target in an individual is important to fully transition to personalized medicine.
Personalized healthcare will result in medication being 3-D printed with carefully tailored dosages, shapes, sizes, and release characteristics to specifically target an individual. Not only will gender, weight, and other physical variables be considered, but drugs could also be designed to specifically target a molecule that may only be present in certain genomes belonging to certain population groups.
Deep learning has successfully being used to predict the binding between a drug and target molecule, this was successfully tested by researchers at the Gwangju Institute of Science and Technology in Korea, as was published in the Journal of Cheminformatics.
A fragmented privatized healthcare system may struggle with the above, but this scenario could be tested today by ambitious politicians interested in moonshot thinking, especially in public healthcare systems.
Estonia is actually leading in this, The Estonian Genome Project is a population-based biological data base and biobank that contains health records from a large percentage of the Estonian population. The Israel Genome Project is another ambitious project to sequence over 100,000 members of the Israeli population.
The clear leader in this space is the UK, there Genomics England collaborates with the NHS to deliver and continually improve genomic testing to help doctors and clinicians diagnose, treat and prevent illnesses, like rare diseases and cancer. They also have a goal of sequencing 100,000 members.
One problem with the above nationalized databases is that the populations sequenced are still relatively small. A larger dataset would result in a higher performance by the AI system when used to identify patterns in the data. More importantly it would result in equal access to healthcare for an entire population.
Also, data from a more diverse ethnic background would be beneficial to prevent potential AI bias issues. AI bias results from data that lacks diversity, including genetic diversity. Canada for example would be best positioned to capitalize on this, due to its status as having one of the largest immigrant populations in the world, with 20% of Canada’s population being foreign born. This contrasts to the Estonian and Israeli systems which suffer from a lack of ethnic diversity.
Standardizing genome sequencing and using deep learning could result in large financial savings and health benefits to the Canadian healthcare system. A system often labeled by Canadians as being broken with long wait times.
The above does require ambitious thinking, regardless any country that undertakes this moonshot thinking would be saving tax payers money, investing in the future, increasing quality of life, and extending the lifespan of its general population – All while leading the world in offering equal access to personalized healthcare and personalized medicine.
Credit: Source link
