Over the last few decades, medical treatment programs have mostly been one-size fits all. Clinicians prescribe different medication or treatment options to patients based on that disease category. However, patients often respond differently to different treatments, so part of the process is working out the most effective treatment with the least severe side effects. Personalized medicine takes an entirely different approach. It aims to develop a precise clinical picture of a patient based on their unique biological makeup. This depth of information allows clinicians to identify variances in genes, RNA or proteins that could affect their susceptibility to a disease.
“By learning the unique disease characteristics specific to each patient or patient subgroup, personalized medicine aims to design smart and tailored drugs that are directly informed by these characteristics to treat disease in a more precise and targeted manner,” says Dr. Ashley Sanders, a group leader at the Max-Delbrück-Centrum für Molekulare Medizin (MDC) in Berlin, Germany.
While personalized medicine is a new approach in modern medicine, the idea of treating patients as individuals is hardly novel. As early as the second millennium BCE, ancient Egyptian medicine, followed later by Greek medicine, focused heavily on understanding an individual’s health, as well as their circumstances and beliefs. The one-size fits all approach arguably came into common usage with the rise of pharmaceutical companies in the early 20th century. Personalized medicine offers a return to these founding principles of treatment.
In this article, we discuss some of the advancements that are bringing personalized medicine into the fore.
Cracking the genetic code
The Human Genome Project was perhaps one of the most important steps towards personalized medicine. In short, it paved the way for medicine to become preventative rather than purely reactive. Published in 2003, the full sequence of the human genome allowed scientists and clinicians to identify the genetic variations of thousands of diseases ‒ an approach known as genomics. It also paved the way for fast, cheap and accurate DNA and RNA sequencing, placing genomics at the core of research and treatment efforts for various human diseases.
Cancer is a great example of how genomics is used to tailor treatments to individual patients based on genetic information. No two patient’s cancers are identical, and how a cancer develops is due to a complex relationship between genes and the environment. DNA sequencing offers a way in, allowing clinicians to identify inherited genes that are known to increase an individual’s susceptibility to certain cancers.
After a diagnosis of cancer, DNA sequencing of tumor biopsies can identify genetic markers of cancer, allowing clinicians to direct certain anticancer treatments towards that specific abnormality (if the drug is available). For example, approximately seven percent of non-small cell lung cancers (NSCLCs) are caused by a heritable mutation in the ALK gene encoding anaplastic lymphoma kinase (ALK). This discovery led to the development of ALK blockers such as ceritinib, which can be selectively prescribed to patients with ALK mutations suffering from NSCLC.
More recently, a new genomics approach called somatic mosaicism is taking personalized medicine one step further to adapt therapies to individual patient cells.
“Our group studies somatic mosaicism, which represents an exciting paradigm shift in how we understand the genetics of disease. We are using single-cell technologies to learn about the molecular changes and DNA mutations that form in our cells and how they change as we age and in disease. We aim to use this information to directly target the mutated cells that may participate or even drive the disease characteristics,” Sanders says.
The rise of multiomics
While genomics was the starting point of personalized medicine, the rise of many other omics technologies is expanding the potential of to avoid repetition. Proteomics and transcriptomics are now used alongside genomics to understand abnormalities that can occur from the genetic code to RNA transcription, to protein synthesis and cellular function.
“These multiomic technologies are central to personalized medicine. It’s what allows us to deeply characterize patients and disease states so that targeted treatments can be developed,” Sanders notes.
Multiomic technologies are used to identify biomarkers of disease much like ALK mutations that occur in a subset of NSCLCs. Proteomics, for example, allows clinicians and scientists to understand all the molecular processes in a patient that drives a disease. Techniques such as mass spectrometry (MS) allow “high throughput” identification and quantification of molecular changes in patients with extraordinary sensitivity and ease.
While older MS-based proteomics had issues with sensitivity and specificity, newer technologies (e.g., liquid chromatography-MS) allow scientists to generate highly detailed data about thousands of proteins from large groups of patients. Indeed, MS is beginning to carve out its own clinical field, MS-based personalized drug therapy, but due to the high costs associated with the approach, only select laboratories can benefit currently. As such, many clinical laboratories continue to use older, low-throughput antibody-based methods for clinical diagnostics until accessibility to MS-based methods are improved.
Big data, big potential
Creating a detailed, holistic view of individual patients requires tremendous computing power. Estimates vary, but as we have between 19,000 and 22,000 genes in our genome, our proteome is roughly the same size. The complexity of understanding how this affects human disease is staggering, and that’s even before you consider variations that arise from alternative splicing, single amino acid polymorphisms and post-translational modifications.
“This is big data which simply can’t be analyzed manually,” says Annalisa Occhipinti, associate professor at Teesside University, UK. “You need advanced computational techniques to find the source of a disease like cancer in all that data,” she adds.
Occhipinti, develops data science tools for use in personalized medicine alongside her colleague Claudio Angione also an associate professor at Teesside University. They have been using data science to predict patient outcomes from multiomics data.
“We start with the patient’s clinical profile and integrate this with multiomics data, and other test results from scans. We bring these data together using machine learning, which links the data from one patient to data from a large cohort of patients. This cross-comparison allow us to predict how the cancer will progress and which drugs could be used to treat that specific cancer subtype in that patient,” Occhipinti says.
The rise of data science has been integral to personalized medicine: it is the node that integrates patient data and determines how to treat their disease. Like easy access to DNA sequencing, readily available data science tools help to integrate multiomics into medical healthcare systems. However, the data science tools that Occhipinti is describing are some way off from clinical use.
“One important question is how accurate the computational predictions are. Is this more accurate than a specialist oncologist? This comes down to training algorithms. The more data that computational models learn from, the more predictive accuracy they have. But perhaps the question is a red herring, as we are working out how data science tools can complement clinician decision making rather than replace it entirely,” adds Angione.
Treating you and your genes
Perhaps ironically, data science is enabling scientists to view patients as more than just a set of numbers. By integrating vast amounts patient data, clinicians and scientists have great potential in understanding how individuals deal with diseases. Afterall, every patient has different ways of coping with and responding to illness.
Going back to the approaches used in ancient Egyptian medicine, personalized medicine must aim to incorporate psychological and environmental aspects of a patient’s life to treat diseases. Many scientists and clinicians are taking this on board, aiming to study subjective aspects of disease states with empirical techniques. For example, scientists have recently identified patterns of brain activity that correspond to clinical pain. While still translational work, a recent study published in Nature Medicine describes the development of an algorithm that can detect chronic pain states in individual patients based on their brain activity recorded with magnetic resonance imaging (fMRI). With this data, clinicians aim to assess if a patient is suffering and what they can do to treat it.
When will personalized medicine reach the clinic?
Tremendous progress has been made to make medicine more personalized since the publication of the human genome in the early 2000s. A boom in multiomics and data science techniques is helping to develop healthcare options that are better tailored to individual patients’ needs rather than fitting them into a one-size fits all approach.
However, according to Sanders, personalized medicine has some way to go before being systematically integrated into healthcare systems: “Generating the multiomics data required to deeply characterize patients requires expensive technologies that are not readily available in clinics. Extracting the disease characteristics from these data requires specialized knowledge and computational experience which takes time to analyze. The current challenge is to reduce the cost and time for diagnosis so that we can implement personalize medicine in routine clinical practice.”
Perhaps what’s needed most to make personalized medicine accessible for more patients is not technological advancements, but methods to streamline personalized healthcare approaches into broader healthcare systems.