We had two representatives from Aigenpulse attending the Festival of Genomics (or FoG as we affectionately call it). As a start-up tech company working at the forefront of analytics and machine learning within the biotech sector, we’re passionate about keeping in tune with current trends (and not to mention that we’re a band of scientists and techies at heart!). So here are some of the things that caught our eye.
A Genomic NHS
This medium-term plan really has the potential to turn NHS England into a world-beater
A packed opening talk was given by Sir Malcolm Grant, Chairman of NHS England, to convey the ambitious aims of the NHS: to become a pioneer in providing a Genomic Medicine Service within 5 years. Driven by initiatives such as the 100,000 genomes project, it aims to provide genomic medicine to 55 million people, as well as (among other goals) to build a new national database that will inform medical research and clinical trials.
While the NHS has traditionally been a slow-moving beast, it looks like this medium-term plan really has the potential to turn NHS England into a world-beater, because of its clearly defined visions and objectives which encompass clear strategic engagement on both clinical and public fronts. There was also a mention of the kind of infrastructure required to provide this operation of personalised/genomic-led medicine, including the need for an advanced, integrated informatics capability that is able to store, secure, connect and distribute patient data, country-wide. Especially in the current climate of cost-saving, and of uncertainty surrounding the privatisation of the NHS, it was inspiring to see this ambitious pursuit that would utilise the latest scientific and technological advancements, with the potential to truly transform the country’s health.
You will find it incredibly difficult to apply fancy algorithms and gain any useful insights from mismanaged data
With the current hot topic being AI/machine learning, a few key points were discussed in the succinctly-named session ‘Integrating Rich Biomedical Data Sets in the Era of Machine Learning & Advanced Analytics’. This session reminded us of some of the basics when it comes to applying these potentially powerful techniques. Chaired by the charismatic and consciously antagonistic Dr. Paul Agapow (Imperial College London), the panel consisted of Bissan Al-Lazikani (Institute of Cancer Research), Mathew Woodwark (MedImmune) and Sofia Olhede (Big Data Institute, UCL).
• ML/AI is not a secret pill pt.1: The old rule of garbage in/garbage out still applies! … and it is probably more important when building predictive models. When it comes to data, both quality and quantity is key.
• ML/AI is not a secret pill pt.2: Biomedical data management is still a huge problem. The fact that this data often uses multiple identifiers, types, formats and silos, coupled with challenges around security, governance and compliance means that you will find it incredibly difficult to apply fancy algorithms and gain any useful insights from mismanaged data.
• Apply the right technology to the right problem – whilst no-SQL or semantic data models are very powerful for certain things, SQL is still perfectly suited to structured and relational data.
• Data standards: Over-standardisation should be resisted! This can turn into an exercise of race to the bottom or identifying the lowest common denominator. Really, the focus should be on data interoperability, especially for clinical data.
(Raising) Public Awareness
There was clear agreement that the adoption of ML/AI in assisting clinical decision making was double-edged sword
The aim of genomic medicine to provide more accurate diagnoses, personalised treatment strategies and a better overall clinical outcome for patients, and with this in mind, the ethics surrounding patient data was also discussed in the above panel session. There was clear agreement that the adoption of ML/AI in assisting clinical decision making was a double-edged sword: that it certainly should be used as a tool in the clinic, but to avoid it being seen as a ‘black-box’ approach to clinical decision-making, there must be: a clear regulatory framework, additional training to staff, and better communication to patients. Further, regarding the usage of patient data, there remains a need for clear governance guidelines (e.g. consent, time limitations on data usage, and the right to an explanation) across the NHS, academic research and industry drug development. The key message was that as a scientific community we are not communicating the benefits of such technologies well enough to the public we serve. As well as this, there remains a clear responsibility for public bodies to take a leading role in building these advanced capabilities, particularly to balance the corporate interest in public health data. Finally, the overarching principle should be firmly placed on the advancement of human flourishing, not only in regulatory and governance design, but also in the public conversation- something with which we wholeheartedly agree.
The most accessible version is outdated, missing genes and other areas that are not correctly arranged
There were 4 AGNC sessions on Wednesday afternoon (AGNC is the Association of Genetic Nurses and Counsellors, for the uninitiated). This is a burgeoning discipline with 310 practitioners in the UK, which acts to support patients and their relatives at risk of or carrying a genetic disorder. I was particularly interested in the talk by Gemma Chandratillake (East of England Genomic Medicine Centre) on “What a Genetic Counsellor should know about a negative result from genomic sequencing”. One inherent problem seems to be that these tests must be compared to a baseline ‘normal’ genome (the Human Reference Genome), and currently, while we have a number of these, none of them are perfect. The most accessible version is outdated, missing genes and other areas that are not correctly arranged. By comparison, the most current version has many of these areas fixed, but has other issues such as nonlinearity, which makes it difficult to interpret. Are the few people who donated their genomic material to make these reference genomes really a good representative of the population at large? These are non-trivial problems to overcome. However, the discipline is rapidly expanding and there is clearly some great talent tackling these challenges.
Blockchain in genomics is starting to look like the new frontier
Blockchain in genomics is starting to look like the new frontier: A secure, unhackable way for individuals to keep their own genetic data, while also being able to release that data in a controlled way to their choice of stakeholders and/or healthcare professionals. They could even to profit from it by selling it for trials etc. There were two blockchain providers at the Festival, Encrypgen & Project Shivom. An article in Forbes talks about the privacy and security of genomic data, which is complicated by the fact that there is no clear, legal owner of genomic data. Encrypgen’s founder Dr David Koepsell has developed a solution based on blockchain technology, combining the values of privacy, security and ownership, while opening up anonymised data for scientific advancement. This is definitely a technology to keep your eyes on.