Hello, I am Jaechang Lim, CTO of HITS.
Lately, many people have been using LLM-based services, haven't they? Their applications are growing not only in daily life but also in research settings. However, in fields like drug discovery that require expertise and precision, existing LLMs have limitations. That’s why HITS has developed an AI assistant specifically for drug discovery researchers.
In this post, I will introduce the capabilities of our AI assistant and how it practically supports researchers in their work.
The AI assistant offers various functions tailored for drug discovery researchers. It can perform tasks such as drug structure and property analysis, activity and experimental analysis, literature search and strategy, target information analysis, competitive drug analysis, and inference.
There are essential but time-consuming tasks in drug discovery that many researchers face. For example, summarizing prior papers on target proteins, comparing similar drugs, organizing and visualizing experimental data, and calculating IC50 are repetitive and time-intensive tasks. These can be automated using AI assistants.
According to tests conducted by our internal researchers, the AI assistant provided more accurate and detailed responses than ChatGPT for a list of 13 drug discovery-related questions. For example, in tasks like finding compounds with similar structures to FDA-approved drugs or calculating solubility and LogP, the AI assistant performed well, whereas ChatGPT did not.
This performance is possible because the assistant’s workflow is designed to execute functions specialized for drug discovery. Instead of simply searching the web, it queries drug discovery-specific databases based on the user’s questions and provides answers grounded in those results. This enhances the depth and specificity of responses, and since it also provides sources, we expect researchers will be able to use it with greater confidence.
Next, I will demonstrate the performance of the HyperLab AI assistant through two specific examples.
First, let’s consider a question about identifying promising drug candidates for a specific target protein (BTK).
“Tell me about the most promising drug among BTK inhibitors.”
In response to this question, the HyperLab AI assistant begins by collecting the latest literature and clinical trial data.
Next, the investigated content is organized into four categories (Next-Generation Covalent BTK Inhibitors, Non-Covalent BTK Inhibitors, Emerging BTK Degraders, and Comparative Advantages), and a conclusion is presented as well. For a drug development researcher, having a comprehensive overview of various approaches and candidate drugs for their target protein would be immensely helpful. Such a task, which typically requires reviewing numerous papers and clinical data, can be accomplished in under 30 seconds with accurate responses using the HyperLab AI assistant.
Beyond the literature review discussed earlier, one of the major strengths of an AI assistant is its ability to visualize user data. When researchers prepare materials, creating graphs from experimental results or writing descriptions for those results can take a surprisingly long time. However, using an AI assistant can significantly reduce this time.
The table below contains concentration-activity data. Based on this data, let’s request the AI assistant as follows: “Analyze the dose-response using the Sample and Activity columns from the uploaded file.”
Upon entering this prompt, the AI assistant will generate a graph based on the data and calculate the IC50 value. A key advantage is that it doesn’t just provide the graph but also includes detailed analysis results, as shown below.
The HyperLab AI assistant is set to evolve, enabling easier and more intuitive operation of HyperLab features through a conversational interface. Our ultimate goal is to help researchers overcome the complexities of drug development and seamlessly integrate AI into their work.
Currently, the HyperLab AI assistant is in beta and available for both paid and free users. We invite you to experience a new level of research efficiency with the HyperLab AI assistant.
HyperLab, AI Drug Development Platform
