With the Bio-IT World Conference and Expo wrapping up last week, it signals the end, for Dotmatics at least, the conference scene in the first half of 2019. For Dotmatics, we now transition to some of our own events, with User-Group Meetings in APAC in Seoul, last week, and this week in Osaka and Tokyo. We continue these events into North America in a couple of weeks, in San Francisco and Boston, and in Melbourne, Australia. So this is a good time to review some of the major news and trends that we saw at the 2019 conferences.
Perhaps pre-eminent amongst these is Artificial Intelligence and Machine Learning (AI / ML). Hype (many, including myself, would say “over-hype”) around AI / ML has grown exponentially over the last couple of years, and, unfortunately, with the increase in activity has come an equally broad re-defining of the terms. At one of the conferences I attended earlier this year, I heard a reference to “AI” that I’m pretty sure wasn’t anything more than some basic statistics! This is undoubtedly an outlier, but even the broad swath of AI / ML discussion bears little-to-no relationship to the original definition of true “AI” – which, in a simplified definition, means an ability for a machine to solve problems that it has not been explicitly coded to do. The concern here is that, per the Gartner Hype Cycle, we will rapidly crest the peak of inflated expectations and fall devastatingly into the trough of disillusionment, and once again the term AI will be persona non grata (if such an attribute can be applied to a term). That would be very unfortunate, because, while true AI may never be observed in our lifetime, ML, the development of which has been driven, at least in part, by efforts to achieve AI, has the potential to provide immense benefits for drug discovery, and many other industries. We know this is true, because it is already a proven technique! Those less acquainted with advanced analytics in drug discovery may not realize it, but computational chemistry approached in drug discovery has been doing ML – e.g., quantitative structure activity relationships (QSAR) – for many years. We are seeing similar approaches now being applied to biologics drug discovery, such as Biological Sequence Activity Relationships (BSAR), championed by Dotmatics, and others.
One promising development, suggesting that the discussion around AI / ML is becoming more grounded, is the recognition that high quality data is essential for proper application of ML techniques. Never has the term “garbage in, garbage out” been more relevant. This provides a nice segue, because the other key topic I’ve heard about at conferences this year is automation. Alongside cycle-time efficiencies, a big benefit of automation is improvements in data quality, from having removed humans from as many steps in the process of generating, capturing, transferring and initial analysis of laboratory data. It’s rather ironic, for while humans of even the meanest intellect can, with no training and little thought, do tasks that the most sophisticated super computers could not even begin to attempt, humans are also devilishly capable of introducing errors into data processing procedures that the simplest circuit board would handle with aplomb*.
Greater levels of lab automation would, therefore, seem to be strongly aligned with genuinely successful applications of ML techniques (I’m dropping the “AI /” part from now on). And to this end, we are seeing a lot of interest. For example, our announcement at SLAS of the integration between Dotmatics and TetraScience engendered huge interest and reflects a significantly underserved need in the industry. The ability to load a plate onto an instrument, start the machine running, and have the next involvement of a human being secondary data analysis (the machine having performed primary analysis) is a compelling use case. At Bio-IT last week we announced the latest release of the Dotmatics Platform – version 5.4, which includes a major enhancement to our Studies assay data management capability – which we refer to as Screening Ultra. One of the key enhancements in Screening Ultra is a much expanded set of Restful APIs to support automation. We’ll be covering these updates in upcoming webinars (5.4 on May 15 and the TetraScience integration on May 23), so please join us if you’re interested to hear how Dotmatics is advancing the cause of automation, and by extension ML, through improvements in data quality, in drug discovery. To learn more about Dotmatics applications of ML per se, check out our product page for Vortex, and of course, feel free to contact us.
* There I go, conferring human attributes to machines in a blog where I complain about the blurring of lines in artificial intelligence…