AI and machine learning has unlocked significant value across the entire healthcare delivery lifecycle.

getty

Medicine has always operated as an “evidence based” field, meaning that it generally pursues experimentation to gather evidence to support a specific claim for diagnostic and treatment success. For example, medications often operate on the laws of statistical averages: if an overwhelming majority of people respond positively to a specific drug, then it is generally prescribed more broadly as applicable to a wider part of the population, until the evidence proves otherwise. However, this also means that a percentage of the population will be impacted by the margin of error and may experience side effects, lack of therapeutic benefit, or perhaps more harmful consequences due to ill-suited therapy.

With the advent of advanced machine learning and deep learning models, this precedent is slowly changing, and drug designers and therapeutics manufacturers are increasingly finding the means to shift towards predictive, genotype-directed care. One key aspect of this pursuit is the study and development of more advanced tools in the field of pharmacogenomics, which is based on understanding an individual’s unique genetic make up and how it may respond to specific medications. A recent study in the Annals of Medicine and Surgery discusses how the incorporation of pharmacogenomics into precision medicine has created massive new opportunities for tailored care; specifically, genotype-guided therapy has become more available due to AI powered modeling of drug-gene interactions in addition to machine learning models being able to synthesize massive volumes of genomic and population level data to determine customized dosing and the propensity for adverse reactions. As the authors indicate, these opportunities are already being leveraged in some of the most critical specialities, including across psychiatry, cardiology, oncology, and infectious diseases. In psychiatry, ML models are increasingly unlocking the ability to predict treatment resistance for antidepressants well in advance, which is a common problem and point of frustration for this patient group; in cardiology, genotype-directed warfarin and clopidogrel dosing have created significant benefits in mortality decreases and adverse event rates. In oncology, perhaps the field that is most prone to success with pharmacogenomics, machine learning models are able to analyze a new granularity of biomarkers, thus enabling “ultra-targeted therapies that strike tumor-specific mutations with remarkable precision.”

This is the future of tailored, bespoke medicine.

Another study published in Frontiers in Artificial Intelligence discusses how the transition of medicine from a one-size-fits-all approach to curated treatments marks one of the biggest shifts in the history of the field; this has largely been made possible by frontier models being able to unlock entirely new ways to analyze and compute data. With regards to disease diagnosis, ML models have created significant opportunities to improve early detection: “Machine learning algorithms have proven to be highly effective in recognizing patterns within complex datasets, allowing for earlier diagnosis of conditions such as cancer, diabetes, and cardiovascular diseases, often before clinical symptoms manifest. For example, AI models have been developed to analyze medical imaging, such as mammograms and CT scans, for signs of malignancies with high accuracy. These models use vast datasets of labeled images to “learn” the features indicative of early-stage cancers, aiding radiologists in identifying potential issues faster than traditional methods. AI’s ability to analyze genomic and clinical data also extends to predictive analytics in assessing patient risk for various diseases. By integrating patient history, lifestyle data, and genetic information, AI algorithms can forecast the likelihood of a disease developing, enabling proactive intervention.”

Why is all of this important?

The frontier AI models that are prevalent today are incredibly powerful. But more importantly, this is the worst that they will ever be. With billions of dollars being poured into AI research, an astounding appetite by both the retail and corporate economy to invest in this technology, and a rapidly deteriorating overall morbidity index globally, the environment is ripe to leverage the best of this technology to truly transform diagnostics and precision care in the coming decades. AI has significant potential to unlock a tremendous amount of value across the entire value chain in healthcare: from diagnosis and detection, to precise drug design, curated therapeutics, remote patient monitoring, post-therapeutic follow-ups and longevity medicine, there is a lot to gain from this technology, if developed correctly.