New drugs typically take 12-14 years to make it to market, with a 2014 report finding that the average cost of getting a new drug to market had ballooned to a whopping $2.6 billion.
It’s a topic I’ve covered before, with a study published earlier this year highlighting how automation could be used to reduce the cost of drug discovery by approximately 70%.
Or you have the team from the University of Toronto, who in a recent paper describe the use of machine learning to generate 3D structures of protein molecules to assist with drug development.
“Designing successful drugs is like solving a puzzle,” the team say. “Without knowing the three-dimensional shape of a protein, it would be like trying to solve that puzzle with a blindfold on.”
The team believe that being able to determine the atomic structure of protein molecules will play a huge role in understanding how they work, and how they may respond to drug therapies. The drugs typically work by binding to a protein molecule, and then changing its shape and thus altering how it works.
Another example of what’s possible is provided by a recent study published in Cell Chemical Biology. The study reveals a big data based approach to detecting toxic side effects that would prohibit a drug from being used on humans before it gets to the expensive clinical trial stage.
Most of these approaches rely on huge datasets to derive their insights, but a recent Stanford University study * highlights the potential of AI, even with relatively small amounts of data to play with.
The team use a new kind of deep learning known as one-shot learning that can do its stuff with relatively few data points.
“We’re trying to use machine learning, especially deep learning, for the early stage of drug design,” the team say. “The issue is, once you have thousands of examples in drug design, you probably already have a successful drug.”
A considerable problem to overcome is the data properties of small molecules. In order to make such data more digestible, the team represented each molecule via the connections that existed between atoms. This process helped them identify a number of properties that could then be analyzed via the algorithm.
The algorithm was trained on two distinct datasets, one on the toxicity of various chemicals, with the other on known side effects from approved medicines. From both datasets, the algorithm was able to predict the toxicity of the medicine with reasonable accuracy.
“We worked on some prototype algorithms and found that, given a few data points, they were able to make predictions that were pretty accurate,” the team say.
On the shoulders of giants
The work is the latest in a series of iterative improvements in one-shot learning. It builds on what has gone before by relying on the closeness of different molecules.
The team believe that their algorithm could be a useful aid to chemists aiming to choose the right molecule to conduct further analysis of in their studies.
“Right now, people make this kind of choice by hunch,” the team say. “This might be a nice compliment to that: an experimentalist’s helper.”
It’s the first time that the one-shot approach has been applied in this way, so is certainly interesting. Suffice to say, it’s very much the start of the journey rather than the end, and it will be fascinating to monitor where it goes from here.
The code itself is openly available via the DeepChem library, so hopefully others will build on it, just as they have built on the work of others.
*study supported by The Hertz Foundation.
Our 2024 Coverage Needs You
It's Another Trump-Biden Showdown — And We Need Your Help
The Future Of Democracy Is At Stake
Our 2024 Coverage Needs You
Your Loyalty Means The World To Us
As Americans head to the polls in 2024, the very future of our country is at stake. At HuffPost, we believe that a free press is critical to creating well-informed voters. That's why our journalism is free for everyone, even though other newsrooms retreat behind expensive paywalls.
Our journalists will continue to cover the twists and turns during this historic presidential election. With your help, we'll bring you hard-hitting investigations, well-researched analysis and timely takes you can't find elsewhere. Reporting in this current political climate is a responsibility we do not take lightly, and we thank you for your support.
Contribute as little as $2 to keep our news free for all.
Can't afford to donate? Support HuffPost by creating a free account and log in while you read.
The 2024 election is heating up, and women's rights, health care, voting rights, and the very future of democracy are all at stake. Donald Trump will face Joe Biden in the most consequential vote of our time. And HuffPost will be there, covering every twist and turn. America's future hangs in the balance. Would you consider contributing to support our journalism and keep it free for all during this critical season?
HuffPost believes news should be accessible to everyone, regardless of their ability to pay for it. We rely on readers like you to help fund our work. Any contribution you can make — even as little as $2 — goes directly toward supporting the impactful journalism that we will continue to produce this year. Thank you for being part of our story.
Can't afford to donate? Support HuffPost by creating a free account and log in while you read.
It's official: Donald Trump will face Joe Biden this fall in the presidential election. As we face the most consequential presidential election of our time, HuffPost is committed to bringing you up-to-date, accurate news about the 2024 race. While other outlets have retreated behind paywalls, you can trust our news will stay free.
But we can't do it without your help. Reader funding is one of the key ways we support our newsroom. Would you consider making a donation to help fund our news during this critical time? Your contributions are vital to supporting a free press.
Contribute as little as $2 to keep our journalism free and accessible to all.
Can't afford to donate? Support HuffPost by creating a free account and log in while you read.
As Americans head to the polls in 2024, the very future of our country is at stake. At HuffPost, we believe that a free press is critical to creating well-informed voters. That's why our journalism is free for everyone, even though other newsrooms retreat behind expensive paywalls.
Our journalists will continue to cover the twists and turns during this historic presidential election. With your help, we'll bring you hard-hitting investigations, well-researched analysis and timely takes you can't find elsewhere. Reporting in this current political climate is a responsibility we do not take lightly, and we thank you for your support.
Contribute as little as $2 to keep our news free for all.
Can't afford to donate? Support HuffPost by creating a free account and log in while you read.
Dear HuffPost Reader
Thank you for your past contribution to HuffPost. We are sincerely grateful for readers like you who help us ensure that we can keep our journalism free for everyone.
The stakes are high this year, and our 2024 coverage could use continued support. Would you consider becoming a regular HuffPost contributor?
Dear HuffPost Reader
Thank you for your past contribution to HuffPost. We are sincerely grateful for readers like you who help us ensure that we can keep our journalism free for everyone.
The stakes are high this year, and our 2024 coverage could use continued support. If circumstances have changed since you last contributed, we hope you'll consider contributing to HuffPost once more.
Support HuffPostAlready contributed? Log in to hide these messages.