This is part 3 of 3 on software for preclinical R&D. Read part 2 here: “Venture Capital's Unfounded Fear of Preclinical R&D Software Hurts Science.”
In 2015, my co-founder Tom Leung was performing cancer research when he experienced first-hand the pain that can result from the lack of preclinical R&D software. A key experiment provided misleading results due to an inappropriate reagent. As a result, Tom lost rare patient samples and his research was set back months. He had spent many hours diligently going through the scientific literature to select this reagent, yet still, it failed him.
Was this Tom’s fault? No. It happens far more often than you might think.
In my first post in this series, I discussed the lack of software in preclinical R&D, and in the second, I explained the reasons for this absence. In the final article of this series, I will illustrate the impacts this has on science, and how we're trying to address them.
Tom's experience is quite common. Such lost time and money in preclinical R&D is a key contributor to pharma's notorious productivity crisis. The pharmaceutical industry has been facing this crisis for over a decade, as R&D costs steadily increase while drug approval rates don’t.
This is a well-known fact within the industry, and efforts have been made to determine the cause. The usual suspects are things like patent expirations causing declines in branded drug sales, increased competition pushing companies into more high-risk high-reward areas of study, and stringent regulatory requirements impeding clinical trials. These all play a part to be sure, but we’re still missing a large piece of the puzzle.
The piece I’m referring to is this: more than 50% of all preclinical experiments don’t provide results that scientifically advance assets. These experiments not only take up valuable scientist time, but they also bloat an organization’s material spend. Experiments are the lifeblood of R&D, so why does this continue to happen?
Most researchers would probably tell you something to the effect of “that’s just the way it is.” Preclinical research is exploratory in nature; thus, a high rate of unproductive experiments is the accepted norm. But it doesn’t have to be this way.
Research has begun to show that, with access to better information, scientists can make better decisions, which in turn leads to a higher rate of experimental success. This being the case, one solution to address challenges like Tom's, and the industry's, is to get the best information into the hands of the scientists that need it. This is much more difficult to achieve in practice than it is in theory.
It isn’t that the information doesn’t exist, far from it. There are millions of experimental data points spread across various online literature. The problems begin to arise when sorting through all that information for the data that is relevant to a specific experimental context. Scientists have to do this manually, which can take weeks, and often the data they find will be incomplete or downright inaccurate.
This is the challenge Tom faced. Tom is a brilliant man, but he’s still only human. If he could comb through every piece of scientific literature in the world, he would have eventually found the data he needed to make his experiment a success. But that could take more than a lifetime. Here is where there's an opportunity for software to make a substantial contribution. A computer has the capability to sort through data much faster than any human ever could. With this assistance, scientists can very quickly locate and compare the data that is relevant for their specific experiments.
Although it’s widely ignored, there are some companies that have recognized and are working to fill the gap in preclinical R&D software. One such company, as I mentioned in the first article of this series, is Benchling. Benchling saw an opportunity to improve scientists’ workflows by transforming ELNs from an out-of-date recordkeeping system to a full suite of software products that augment scientists’ workflow and elevate their productivity. They now provide solutions for scientists researching antibodies, cell therapy, gene therapy, proteins and peptides, strain engineering, and vaccines.
Two more companies making headway in the preclinical software environment are Scientist.com and Science Exchange. These companies provide online marketplaces for research services, allowing scientists to outsource aspects of their R&D to academic institutions, CROs, and other scientific organizations. The model facilitates a faster, leaner approach to research and promotes collaboration and innovation to expedite scientific advancement.
At BenchSci, we have seen the impact of software on experimental success first-hand. We use our technology to identify and resolve Avoidable Experiment Expenditure (AEE), which refers to all inefficiencies in preclinical R&D. The financial impact of AEE on the pharmaceutical industry is significant, amounting to nearly $48 billion lost annually worldwide.
Our current focus is on the reagent selection process, which is a significant contributor to AEE. Over a third of AEE (totalling over $17 billion worldwide) can be attributed to inappropriate reagents, including reagents that don’t work in the context in which they are used, as well as reagents that may work to an extent but are not the most optimal.
Using proprietary machine learning algorithms, we developed our AI-Assisted Reagent Selection platform. We trained a computer to think like a Ph.D. biologist, enabling scientists to make better decisions when selecting reagents. Our first iteration focused on antibodies specifically, as they are the most commonly used reagents. We are now working on expanding our reach to all reagents, and have already added recombinant proteins, RNAi, and CRISPR to the platform. To date, the AI has decoded over 10.8 million scientific articles and 18.1 million data points, as well as over 28.5 million reagent products from more than 261 vendors.
BenchSci accelerates pipelines and reduces material spend by facilitating better experimental design for increased experimental success. The more scientists in an organization that adopt our reagent selection platform, and the more aspects of experimental design we can cover (additional reagent types, for example), the greater impact we are able to make on that organization’s ROI.
As companies like Benchling, Scientist.com, Science Exchange, and BenchSci show, inefficiencies in preclinical R&D that affect pharma's productivity aren't inevitable. Increasing evidence shows that software can improve workflows and decisions. At BenchSci, we have quantified this within our customers, reinforcing our belief in the value of software for life sciences. We strongly believe that, despite headwinds against investments in preclinical R&D software, pharma companies and venture capitalists should take notice of the opportunities and changes underway.
To learn more about the problem we solve, and how we're solving it with AI-powered software, please download this whitepaper: Avoidable Experiment Expenditure: Examining a major issue affecting life science productivity.