For years, Sebastian Misztal suffered with severe haemophilia, an incurable, genetically inherited blood disorder where the body’s ability to create blood clots is impaired. For patients like Misztal, this means a minor shaving cut would bleed for hours, while slightly bigger injuries necessitate a trip to the emergency room.
Today, scientists and doctors have managed to mitigate many of the symptoms of Misztal’s haemophilia through gene therapy, dramatically improving the quality of life for him and other patients suffering with the same disease. Similar genomic treatments are being developed to fight cancer, Parkinson’s, AIDS, Alzheimer’s, and cardiovascular diseases, and help ease the lives of millions worldwide. All of this has been made possible by the big strides made in gene therapy over the past decade.
Gene therapy is a technique that treats the disease and improves immunity by modifying, adding, deactivating, or replacing the genes within the human DNA. It can be administered both inside and outside the body, where viral vectors are used as vehicles to carry the new genetic material. Typically, pills and other drugs only act as a workaround to treat the symptoms rather than the actual ailment. Gene therapy on the other hand has proven to be highly effective against diseases as it targets the source or cause of the disease.
How can machine learning help gene therapists?
Despite being a revolutionary treatment methodology, the success of gene therapy lies in ensuring the therapeutic gene is specifically targeted to the right cells and the right tissue. If poorly executed, gene therapy could potentially encourage harmful mutations in the DNA, resulting in bigger health problems for the patient. In addition to cell targeting, other challenges facing gene therapists include inadvertently incorporating the gene into germline cells, as well as accurately assessing the body’s natural immune responses to viral vectors that carry the gene-modifying DNA payloads.
A key solution to address all of the above challenges is genome sequencing, a technique that breaks down and sequences the various components of a DNA to better understand their function within the overall genetic code. However, human DNA is made up of over 3 billion such components, and analyzing this wealth of information through traditional statistical techniques can be slow, resource-intensive, and cumbersome.
This is where Machine Learning (ML) systems are proving to be invaluable. ML is one of the techniques used in Artificial Intelligence (AI) to produce automated intelligent behavior by using statistical data. It provides a learning opportunity from statistical data and performing actions based on the learnings. By using ML, the analysis of enormous and complex data can be achieved in a faster and effective manner.
Here are a few value propositions that ML offers to gene therapy:
1. Genome Sequencing
ML accelerates analysis of the sequenced data and also effectively predicts the genetic alterations associated with a particular disease. This helps minimize the time and effort required in the process of developing precision medicine. Algorithms are designed based on patterns identified in large data sets, which are translated to human models to understand their impact.
2. Gene Editing
There are now technologies like CRISPR that can alter DNA sequences to correct gene defects and treat diseases. However, there are chances of off-target effects leading to mutations during the process. ML has the capability to predict the mutations that CRIPSR can introduce into a cell. This could lead to safe and effective CRISPR treatments.
3. Clinical Workflow Optimization
ML helps in better accessibility of patient data, thereby increasing the efficiency in the management of clinical workflow processes. Certain diseases like haemorrhage require immediate diagnosis to increase the chances of successful clinical outcomes. In such cases, ML algorithms can reduce the time to recognize the haemorrhage and many a times could help neuro-radiologists identify the false positives as well.
4. Predictive Genetic Testing & Preventive Medicine
Newborn genetic screening is becoming a standard practice in recent times. Non-invasive screening for diseases such as Down syndrome is being made available to women during pregnancy. The ability of ML to predict outcomes and the risks involved in curing such diseases based on the available data, can aid significantly in performing the genome analysis of a human body.
Over the next 10 years, ML algorithms and their predictions will help map the complex association between risks, such as the occurrence of single nucleotide polymorphisms (SNP) and complex disease phenotypes. This will lead to devising more and more gene therapies for life-threatening diseases and genetic ailments. The success of gene therapy will mean fewer medicines, fewer visits to physicians over the lifetime of a patient, which in turn will improve the Quality Adjusted Life Years (QALY) post therapy. ML-assisted gene therapy promises to be a ray of hope for patients with rare genetic diseases who currently cannot be cured using existing drugs and treatments..