arXiv subjects

Peilin Kang

Publications and source records attributed to Peilin Kang.

6 recordsLinked to original sources

arXiv2026

Navigating committor landscape of biomolecules with a general pairwise interaction model

Sampling rare conformation transitions between metastable states is a central challenge in atomistic simulations. While the committor function serve as an ideal reaction coordinate for driving enhanced sampling, their high-dimensional inputs and complex functional forms limit the efficacy of standard feedforward neural networks in modeling them. Inspired by recent breakthroughs in biomolecular structure prediction, we propose a novel committor learning framework grounded in the AlphaFold 3 paradigm. By integrating a lightweight, differentiable atom-level embedding with a simplified Pairformer architecture, our method inherently captures intricate dynamical features of diverse biosystems without requiring specialized prior knowledge. We demonstrate the superior expressiveness and accuracy of the proposed framework across multiple atomistic processes. For the folding of the chignolin mini-protein, our model reveals the finer-grained structure of its transition state ensemble (TSE) and a detailed bifurcated reaction mechanism. Furthermore, for calixarene host-guest systems, we develop a unified committor model that elucidates how ligand substituents regulate the ratio between distinct binding pathways, offering new perspectives for structure-based drug design.

physics.comp-ph↗

arXiv2025

Committors without Descriptors

The study of rare events is one of the major challenges in atomistic simulations, and several enhanced sampling methods towards its solution have been proposed. Recently, it has been suggested that the use of the committor, which provides a precise formal description of rare events, could be of use in this context. We have recently followed up on this suggestion and proposed a committor-based method that promotes frequent transitions between the metastable states of the system and allows extensive sampling of the process transition state ensemble. One of the strengths of our approach is being self-consistent and semi-automatic, exploiting a variational criterion to iteratively optimize a neural-network-based parametrization of the committor, which uses a set of physical descriptors as input. Here, we further automate this procedure by combining our previous method with the expressive power of graph neural networks, which can directly process atomic coordinates rather than descriptors. Besides applications on benchmark systems, we highlight the advantages of a graph-based approach in describing the role of solvent molecules in systems, such as ion pair dissociation or ligand binding.

physics.comp-ph↗

arXiv2025

The role of fluctuations in the nucleation process

The emergence upon cooling of an ordered solid phase from a liquid is a remarkable example of self-assembly, which has also major practical relevance. Here, we use a recently developed committor-based enhanced sampling method [Kang et al., Nat. Comput. Sci. 4, 451-460 (2024); Trizio et al., Nat. Comput. Sci. 1-10 (2025)] to explore the crystallization transition in a Lennard-Jones fluid, using Kolmogorov's variational principle. In particular, we take advantage of the properties of our sampling method to harness a large number of configurations from the transition state ensemble. From this wealth of data, we achieve precise localization of the transition state region, revealing a nucleation pathway that deviates from idealized spherical growth assumptions. Furthermore, we take advantage of the probabilistic nature of the committor to detect and analyze the fluctuations that lead to nucleation. Our study nuances classical nucleation theory by showing that the growing nucleus has a complex structure, consisting of a solid core surrounded by an interface that is more disordered than bulk liquid. We also compute from the Kolmogorov's principle a nucleation rate that is consistent with the experimental results at variance with previous computational estimates.

cond-mat.stat-mech↗

arXiv2025

Everything everywhere all at once: a probability-based enhanced sampling approach to rare events

The problem of studying rare events is central to many areas of computer simulations. In a recent paper [Kang, P., et al., Nat. Comput. Sci. 4, 451-460, 2024], we have shown that a powerful way of solving this problem passes through the computation of the committor function, and we have demonstrated how the committor can be iteratively computed in a variational way and the transition state ensemble efficiently sampled. Here, we greatly ameliorate this procedure by combining it with a metadynamics-like enhanced sampling approach in which a logarithmic function of the committor is used as a collective variable. This integrated procedure leads to an accurate and balanced sampling of the free energy surface in which transition states and metastable basins are studied with the same thoroughness. We also show that our approach can be used in cases in which competing reactive paths are possible and intermediate metastable are encountered. In addition, we demonstrate how physical insights can be obtained from the optimized committor model and the sampled data, thus providing a full characterization of the rare event under study. We ascribe the success of this approach to the use of a probability-based description of rare events.

physics.comp-ph↗

arXiv2024

Computing the Committor with the Committor: an Anatomy of the Transition State Ensemble

Determining the kinetic bottlenecks that make transitions between metastable states difficult is key to understanding important physical problems like crystallization, chemical reactions, or protein folding. In all these phenomena, the system spends a considerable amount of time in one metastable state before making a rare but important transition to a new state. The rarity of these events makes their direct simulation challenging, if not impossible. We propose a method to explore the distribution of configurations that the system passes as it translocates from one metastable basin to another. We shall refer to this set of configurations as the transition state ensemble. We base our method on the committor function and the variational principle to which it obeys. We find the minimum of the variational principle via a self-consistent procedure that does not require any input besides the knowledge of the initial and final state. Right from the start, our procedure focuses on sampling the transition state ensemble and allows harnessing a large number of such configurations. With the help of the variational principle, we perform a detailed analysis of the transition state ensemble, ranking quantitatively the degrees of freedom mostly involved in the transition and opening the way for a systematic approach for the interpretation of simulation results and the construction of collective variables.

physics.comp-ph↗

arXiv2021

Understanding Catastrophic Overfitting in Adversarial Training

Recently, FGSM adversarial training is found to be able to train a robust model which is comparable to the one trained by PGD but an order of magnitude faster. However, there is a failure mode called catastrophic overfitting (CO) that the classifier loses its robustness suddenly during the training and hardly recovers by itself. In this paper, we find CO is not only limited to FGSM, but also happens in $\mbox{DF}^{\infty}$-1 adversarial training. Then, we analyze the geometric properties for both FGSM and $\mbox{DF}^{\infty}$-1 and find they have totally different decision boundaries after CO. For FGSM, a new decision boundary is generated along the direction of perturbation and makes the small perturbation more effective than the large one. While for $\mbox{DF}^{\infty}$-1, there is no new decision boundary generated along the direction of perturbation, instead the perturbation generated by $\mbox{DF}^{\infty}$-1 becomes smaller after CO and thus loses its effectiveness. We also experimentally analyze three hypotheses on potential factors causing CO. And then based on the empirical analysis, we modify the RS-FGSM by not projecting perturbation back to the $l_\infty$ ball. By this small modification, we could achieve $47.56 \pm 0.37\% $ PGD-50-10 accuracy on CIFAR10 with $ε=8/255$ in contrast to $43.57 \pm 0.30\% $ by RS-FGSM and also further extend the working range of $ε$ from 8/255 to 11/255 on CIFAR10 without CO occurring.

cs.LG↗