Artificial Intelligence and its usage in "Pharma Universe"
What is Artificial Intelligence (AI)?
Intelligence is the ability to detect patterns, store these patterns as knowledge and then use this knowledge to draw conclusions. These patterns look like different layers and they can be processed in a hierarchical away to derive conclusion.
Case for “Artificial Intelligence”
Although AI helps in a big way in faster decision making on complex problems by identifying through neural networks, which acts as inputs, the inputs then get processed which finally help us in taking a decision.
A simple and live example from our everyday life is google translation , we input characters in English or any of our preferred language and we get output in any desired language, how does it work?, The computer recognizes these characters or word/s, and then looks for corresponding words in the memory. The computer repeats this action word by word and then gives us the corresponding output in the desired language, however since computer don’t have human intelligence, the meaning that the computer throws may or may not be mostly be perfect and will have certain limitations. Computer can just match word by word, which may or may not make sense from meaning perspective. Whereas we as humans can learn, abstract and change our behavior basis situations and derive very precise conclusions.
Hence AI (Artificial Intelligence) is good but it comes with limitations, computer will be able to translate at a faster pace but it will be difficult for computer to be able to intuitively throw decisions, that is why computers are very good in playing chess or identifying new patterns but can’t replace human intelligence completely. However the AI’s validity will definitely increase bit by bit with more usage and more data going as input to help draw the most accurate pattern.
Google car, Google Search, Google Translation is all very relevant cases of artificial intelligence
Artificial intelligence and Pharma Universe
In Pharma’s context, what is that we can really do to use AI for certain processes?
In pharma’s context it all started with a Big hype – "In the initial screening, the pharma researchers in the industry had certain molecule targets that were identified, the computer was then made to or was expected to set models of molecules to fit the targets , the idea was to do drug discovery through AI, which was to identify the best molecule that puts molecule to the target, easy isn’t it?
But in real world it’s not - the likelihood of a molecule identified through the AI route to hit the target looks next to impossible, this is because, in true biology because this simply doesn’t work. We as humans were needed to run our own experiments to discover the molecule which the computer can’t do, with AI an attempt was made to use AI for creating shortcuts in the Pharma space which didn’t succeed much because of the necessity of Human Intervention.
Another try was to go to limited available databases and find certain articles and certain problems to identify better molecules and better targets , with limited help , computer could have identified and projected few groups but without access to true and relevant findings it was difficult for computer to predict, complete pattern identification.
What the industry learnt from those experiences was that, we can do sequential layering of certain things and then by incrementally applying these, we can solve low level operation tasks, please note just operational and predictive tasks and not where human intelligence are needed.
What the pharma industry learnt from these early experiments was that Predication of medically different casualty can identify signals, delineate, risks and validate hypothesis and that’s where it stops.
Where can the Pharma world benefit most from Artificial INTELLIGENCE?
PV ( Pharma-co-vigilance)
AI is helping the PV teams globally in a big way, in case management (Adverse Events) , in clinical trials and in certain patterns in different cases, for ex in case management, if we get a text message saying I got a headache after consuming “XYZ” medicine which is like extracting certain inputs, by this alone there is a correlation – “XYZ” and headache , how is this AE related to other similar cases, “XYZ” and pain in head.
Through the AI route computer will start detecting similar past instances and come back with resolutions/alternate resolutions or how was this handled in the past and what is the best way to handle in the current context.
Clinical Trails
Potential Fraud detection is clinical trials is more like the credit card fraud, where the bank detects any suspicious transaction and raise an alarm to the user to check the genuineness of the transaction. Similarly in clinical the use of AI is to focus more on the quality of “Clinical Trails”, for ex is there any fraud or is there any manipulation in the results, AI can help raise an alarm. This is very important, as in the clinical trials process there is usage of lots of data and during the process there are multiple issues - data accuracy, consistency of the data, is the data real or forged?
AI helps in all of these in a big way, simply because mathematic signals are used to compare and see the range of patterns and check the veracity. In cases of clinical frauds companies invents patients and come out with pre meditated patters which is very different in real life and the outliers are seen as a signal and used for fraud detection.
Medical Monitoring
Mathematical data can suggest real time that how do they trial go? How the patients behave? Does reality meets my expectations? Vetting the data to have the analysis can help fasten the process, but the only thing it can’t fasten is the human part of drug taking and absorption.
In the right context, the dialogue between humans and computer which is “Repeatable” and “Trainable” and can never be eliminated by AI, simply put computers can’t do what humans can do which is to define the problem and convert the input, but when it comes to execution with defined commands computers can do better than humans and that is where they should be used the most.
Arjun Singh – Mumbai – 29th March 2017
*Views expressed are personal
Credits and Relevant Reads:
AI | Analytics | Visualization
6yRightly said. AI has a lot to offer as we have yet to master it. But with time as we might enter the phase of Artificial General Intelligence, we will witness something that we couldn't even imagine of. AI in Pharma has life-saving capabilities which the world desperately needs.
Good read Arjun Singh
Technical Consultant at 'Self-Employed'
7yI agree with the general cautionary tone of this article about use of AI in the drug discovery programmes. However, AI could still surprise us. With increasing data processing capacities , it may be possible to predict drug candidates with little more accuracy. However, the key question is whether massive processing can be equated with intelligence. AI paradigm itself needs to be revisited. AI does the same things that we humans do but not by the same methods. The output may be same but the process is not. We need to evolve truly intelligent machines before claiming the label of AI! Dr. Pradeep Chhaya.
Leadership Role- Global Purchasing (Strategic Sourcing, Category Management, SCM & SQE) | Vesuvius | Metso:Outotec | Arceormittal | Tata Steel |
7ynice article...