Enhancing Accuracy of Precision Medicine Through Mathematics
Development of Single Sequencing Lens to Eliminate Noise from Transcriptome Data

As societies worldwide rapidly transition into super-aged demographics, the rising cost of healthcare has brought renewed attention to the importance of precision medicine. Tailored approaches to disease management and treatment are becoming essential. Recognizing this trend, South Korea designated the development of precision medicine technology as a strategic national science and technology project in 2016, focusing on utilizing bioinformatics. Other countries, including the United States, Japan, and China, have also launched precision medicine initiatives with significant investments.

Precision medicine leverages genetic, disease, and lifestyle data to propose effective, personalized treatment plans. With the rapid growth of artificial intelligence and big data technologies, the precision medicine market is poised to expand further, spurring active research and development in this field.

In April, the Biomedical Mathematics Group (IBS BIMAG) within the Institute for Basic Science developed a tool called Single Sequencing Lens (scLENS) to enhance the accuracy of single-cell transcriptome data. This breakthrough has garnered significant attention in the field.

A transcriptome represents all RNA produced in a cell or tissue at a given time, providing crucial insights into drug response prediction, disease diagnosis, and gene expression patterns. Single-cell transcriptome analysis has applications in various areas of precision medicine, including targeting cells with genetic mutations, drug development, and clinical research.

However, analyzing single-cell transcriptome data is challenging due to the sheer volume of information—reaching tens of gigabytes—and the significant noise contained in the data, which often leads to errors. scLENS automatically distinguishes signal from noise in single-cell transcriptome analysis without requiring subjective user adjustments. Since its announcement in April, this tool has been hailed for its potential to secure more accurate genetic data.

KIM Hyun, the senior researcher at BIMAG who participated in the study, explained, “Among the vast amount of data, biologically meaningful signals constitute only about 3%. Traditional methods relied on researchers manually setting the boundaries between biological signals and noise, leading to inconsistent results. Our newly developed technology addresses this limitation, allowing for more reliable data analysis.”

After joining BIMAG in October 2021, Kim Hyun led the development of scLENS. He shared the significance of this achievement, the challenges faced during its development, and his future research plans in this interview.

Q: Please introduce yourself.
Hello, I am KIM Hyun, a senior researcher in the Biomedical Mathematics Group at IBS. I graduated from the Department of Physics at Korea University, where I also earned my Ph.D. in nonlinear dynamics and biophysics. My academic journey into physics started by chance, but the more I studied, the more I enjoyed it. During my doctoral studies, I worked under Professor LEE Kyungjin, analyzing the collective movements of cancer tissues. I also collaborated with Dr. MIN Cheol-Hong at the Catholic University to use mathematical models for studying synaptic connectivity in the brain's SCN. My connection with CI KIM Jae Kyoung, who leads our group, began through a seminar on brain-related research and led to my joining IBS in 2021 as a postdoctoral researcher. At BIMAG, I study biomedical multiscale dynamics with CI Kim.

Q: Please tell us about the Biomedical Mathematics Group.
The Biomedical Mathematics Group (IBS BIMAG) is part of the IBS Center for Mathematical and Computational Sciences. Established in March 2021, BIMAG aims to conduct world-class research in mathematical biology. Our group focuses on simplifying complex multiscale systems, applying these methods to study enzyme reactions, immunotherapy, and the molecular mechanisms of circadian rhythms. We also develop theories to infer dynamic intercellular networks, using them to uncover the cellular network of biological clocks. Additionally, we work on digital therapeutics for sleep disorders based on data from wearable devices like smartwatches.

Q: What inspired you to start this research, and what challenges did you face?
The idea for scLENS began with the goal of creating a tool that could extract biological information more accurately and reliably. We also wanted to make it easier for bioinformatics researchers to use through automation. We started the research in 2022, and it took about two years to complete. One of the challenges was studying and applying mathematical techniques for data analysis. Another difficulty was explaining the improved algorithm clearly. Transcriptome data is both vast and noisy. While certain genes are highly expressed, the majority are weakly expressed. If we simply use the data as is, the information from most genes cannot be interpreted, and only certain gene data ends up representing the overall features. Achieving balance was necessary, but the preprocessing stage was complex.

CI Jae Kyoung Kim advised simplifying the problem first, which was helpful. At the time, I had two young children. Even while cleaning bottles for my second child, I was constantly thinking about the problem, but it was frustrating when no ideas came to mind. Then, one day, while driving with my family, the solution suddenly came to me. I immediately pulled over, switched seats with my wife, and started jotting down notes. When I explained the idea to CI Kim, he approved it right away. Looking back, the solution seems simple, but it was challenging to reach at the time.

Q: What has been the reaction to your results.
scLENS has shown superior performance compared to other programs. When compared with the widely used Seurat program, scLENS achieved over 10% better cell grouping performance and captured 43% more of the local structure inherent in the data. Despite conducting more calculations than other programs, scLENS is optimized for memory usage, enabling it to analyze large-scale data sets—100,000 cells and 20,000 genes—in just three hours. Since the publication of our paper, citations are steadily increasing, and the tool has generated significant interest at conferences. We are also in the process of securing a patent.

Q: What are your future plans?
Our tool extracts signals from data, allowing for diverse analyses. For example, we can cluster cells by type, investigate their interactions, or explore communication between different cell types. We’ve made scLENS available on GitHub as open-source software, enabling anyone to use it. We’re also working to improve its user-friendliness. The more people use it, the more accurate the results will become.

Q: Any final words you’d like to share?
I’d like to emphasize the importance of not limiting yourself. Staying in one place for too long can narrow your perspective and lead to self-imposed limits. Joining IBS exposed me to new environments, standards, and methods, broadening my horizons. My research has also expanded through collaborations with researchers here and abroad. I encourage others to experience new environments—it’s a valuable way to grow.