Earlier this year, we announced the acquisition of Vyasa by Certara, Inc. （纳斯达克股票代码：CERT) a global leader in biosimulation and other software solutions helping the entire drug development lifecycle.
I’m very excited about this opportunity for several reasons, but most importantly the opportunities that this partnership provides for Vyasa and Certara’s clients.
The life sciences industry faces a common challenge – organizations have massive amounts of valuable data that is saved in a variety of formats from structured spreadsheets filled with drug profiles to unstructured clinical study reports to small compound SMILES strings. Historically this data has been difficult to integrate for analysis, but new approaches to deep learning AI are revolutionizing how organizations approach and gain value from their content.
A New Path for Analytics
We started Vyasa in 2017 with a vision for deep learning and a belief that technology was at a tipping point toward advancing from experimental projects to applications that would enable the acceleration of drug discovery and development.
Around this time were a number of exciting developments that presented a “perfect storm” for deep learning – the onset of big data, availability of relatively cheap, accessible storage and the release of powerful computing infrastructures, in particular graphical processing units (GPUs).
Traditional models relied heavily on structured data, but as GPUs have proliferated the market, data scientists have been able to stretch what training data is used for A.I. models. Most notably, unstructured content can now be used as training data.
These advanced deep learning models, known as large language models (LLMs), were first introduced approximately 5 years ago. Since then, numerous other models have been released. For example, NVIDIA’s Megatron-LM and OpenAI’s GPT3 and most recently ChatGPT.
What differentiates these models from traditional deep learning recurrent neural networks is how they train on data sets. Unlike a traditional model that processes each term separately and outside the context of the sequence, LLMs use self-attention to build rich representations of each constituent in the data span, allowing these models to understand the relevance of the location of a term, the relation of one term to the next (even if far away from each other) and more. When trained on larger datasets, these models reach remarkable accuracy and recall for understanding unstructured data like large volumes of natural language text.
These capabilities present a tremendous opportunity for how we interact with large repositories of data which is particularly relevant when considering several use cases across drug development.
Applying Deep Learning to the Drug Development Lifecycle
Certara has spent decades developing a robust suite of tools and services that help life science companies quickly analyze and make sense of data needed to inform drug development decision making. By applying Vyasa’s expertise in deep learning across Certara’s product portfolio, we’re developing exciting new ways for our customers to expand how they approach their data from early-stage compound discovery all the way to medical writing and market access. For example, deep learning is enabling our customers to:
- Improve analysis of scientific literature such as clinical trials and pre-prints through AI-powered concept tagging and named entity recognition.
- Conduct large-scale document insight extraction and create structured spreadsheets for reporting and regulatory review.
- Enhance small compound analysis and enable de novo compound generation for new drugs.
- Standardize datasets across multiple fields and documents to improve clinical trial design and enrollment.
- Accelerate medical writing and patient narratives.
Our Next Chapter
In the weeks since the acquisition, numerous exciting synergies between Vyasa and Certara are evident. We’re excited to soon begin to share some new integrations between our products and continue to see some cool new deep learning applications for Layar as well.
The past six years building Vyasa have been a tremendous experience – and I’d be remiss if I didn’t thank the whole Vyasa team for their work in helping to get us to this exciting milestone. We’ve experimented, we’ve iterated, we’ve pivoted, and, in the end, we’ve established some incredible ways for deep learning technologies to help provide real impact in life sciences.
Interested in learning more? Contact us for a demo or visit the Vyasa product page.
