기초과학연구원 의생명수학그룹
In this talk, we discuss the paper "N-BEATS: Neural basis expansion analysis for interpretable time series forecasting" by B. Oreshkin et al., ICLR, 2020. Abstract We focus on solving the univariate times series point forecasting problem using deep learning. We propose a deep neural architecture based on backward and forward residual links and a very …
Classical machine learning models are typically trained under the assumption that the training (source) and test (target) data are drawn from the same distribution. However, real-world data are rarely clean or consistent, and distribution shifts between the source and target domains are ubiquitous. Despite its importance, addressing distribution shifts is highly difficult. The fundamental challenge …
Recent advances in data science have expanded the scope of data analysis beyond prediction accuracy toward interpretability, causal understanding, and generalizable learning across complex data structures. This lecture introduces three emerging methodological approaches that can be directly leveraged in modern data analysis workflows. First, the lecture presents explainable artificial intelligence (XAI) techniques, focusing on SHAP …
Mathematical modeling provides essential quantitative insights that accelerate drug and cell therapy development. In this presentation, we utilize kinetic frameworks to optimize the design of molecular glues by elucidating their biophysical determinants and identify a key target for NK cell-mediated immunotherapy through systematic data analysis. Collectively, we demonstrate how mathematical strategies can effectively guide and …
In this talk, we discuss the paper "Seasonal timing and interindividual differences in shiftwork adaptation" by R. Kim et al., npj digital medicine, 2025. Abstract Millions of shift workers in the U.S. face an increased risk of depression, cancer, and metabolic disease, yet individual responses to shift work vary widely. We find that a conserved …
In this talk, we discuss the paper "scPPDM: A Diffusion Model for Single-Cell Drug-Response Prediction" by Z. Liang et al., arxiv, 2025. Abstract This paper introduces the Single-Cell Perturbation Prediction Diffusion Model (scPPDM), the first diffusion-based framework for single-cell drug-response prediction from scRNA-seq data. scPPDM couples two condition channels, pre-perturbation state and drug with dose, …
Complex diseases, such as cancer, sarcopenia, and immune disorders, arise from abnormalities in multiple genes and pathways, posing significant challenges to conventional single-target drug discovery strategies. To address this, we developed a perturbome-based analytical framework that integrates transcriptomic signatures, network pharmacology, and machine learning to identify effective therapeutic candidates. Central to this approach is the …
In this talk, we discuss the paper, "Quantifying interventional causality by knockoff operation" by Xinyan Zhang and Luonan Chen, Science Advances, 2025. Abstract Causal inference between measured variables is crucial to understand the underlying mechanism of complex biological processes at a network level but remains challenging in computational biology. We propose an innovative causal criterion, …
In this talk, we discuss the paper, "A wearable-based aging clock associates with disease and behavior" by A. C. Miller et al., Nature Comm, 2025. Abstract Aging biomarkers play a vital role in understanding longevity, with the potential to improve clinical decisions and interventions. Existing aging clocks typically use blood, vitals, or imaging collected in …
In this talk, we discuss the paper "Generic Temperature Response of Large Biochemical Networks" by Julian B. Voits and Ulrich S. Schwarz, PRX Life, 2025. Abstract Biological systems are remarkably susceptible to relatively small temperature changes. The most obvious example is fever, when a modest rise in body temperature of only few Kelvin has strong …
In this talk, we discuss the paper "Multi-Marginal Flow Matching with Adversarially Learnt Interpolants" by O. Kviman et al., 2025, arxiv. Abstract Learning the dynamics of a process given sampled observations at several time points is an important but difficult task in many scientific applications. When no ground-truth trajectories are available, but one has only …
Abstract (국문) 최근 거대언어모델(LLM)을 기술의 발전은 AI4Science 분야에서 Foundation Model 개발에 대한 세계적인 관심을 촉발하였다. 그 중에서도 신약 및 신소재 개발에 연계된 Molecular 도메인에서의 Foundation Model 연구는 막대한 산업적 영향력과 가치를 가지고 있다. 본 발표에서는 분자 구조 생성, 물성, 및 반응 예측 문제에 적용되기 위해 필요한 Multimodal LLM 연구 성과와 방향성을 소개하고자 한다. (English) The …