3986089

Tensor-train based approach for building Markov State Models and Generalized Master Equation Models to study biomolecular dynamics | Poster Board #2144

Date
March 19, 2024
Explore related products in the following collection:

In the recent years, the data-driven analysis of dynamical systems based on Koopman operator have achieved great advances. In addition to neural network based approaches, tensor-train like the multi-linear version of the AMUSE algorithm (AMUSEt) can provide an alternative approach to do analysis of large scale systems with high-dimensional features. Previously, AMUSEt has been applied on large scale systems like NTL9 by manually choosing the features like pairwise distances. However, when applied to other biomolecules, the performance of AMUSEt will heavily rely on the choice of input features. Hence in this work, we first introduce TICA to improve AMUSEt. TICA can provide a good set of input features for AMUSEt. The performance of TICA-AMUSEt is even better than TICA for NTL9 and FIP35 WW domain. This may be related to the overlapping between TICs based Gaussian basis. The performance of TICA is usually limited by the linear combination of features that has no overlapping between TICs; however, the Gaussian basis in AMUSEt can introduce overlapping between input features that allows a second chance to find slow dynamics. In FIP35 WW domain, we show that the performance of AMUSEt is equivalent to VAMPnets. In addition, we also combined the generalized master equation (GME) with the TICA-AMUSEt, showing that GME is able to further improve the TICA-AMUSEt method by accurately computing the timescales of long-time dynamics.

Presenter

Speakers

Speaker Image for Cecilia Clementi
Rice University
Speaker Image for Xuhui Huang
University of Wisconsin-Madison

Related Products

Thumbnail for Efficient path classification algorithm based on Variational Auto-Encoder to identify metastable path channels for complex conformational changes | Poster Board #2117
Efficient path classification algorithm based on Variational Auto-Encoder to identify metastable path channels for complex conformational changes | Poster Board #2117
Conformational changes (i.e., dynamic transitions between pairs of conformational states) play important roles in many chemical and biological processes…
Thumbnail for Unraveling the role of bacterial RNA Polymerase trigger loop dynamics in antibiotics inhibition using neural network and generalized master equation-based models | Poster Board #1621
Unraveling the role of bacterial RNA Polymerase trigger loop dynamics in antibiotics inhibition using neural network and generalized master equation-based models | Poster Board #1621
Bacterial RNA Polymerase (RNAP) is a vital enzyme that transcribes DNA to RNA and is an effective target for antibiotics. Trigger Loop (TL) and Sequence Insertion 3 (SI3) are two domains of RNAP which undergo conformational change during transcription elongation…
Thumbnail for Developing novel ligands with enhanced binding affinity for the sphingosine 1-phosphate receptor 1 using machine learning
Developing novel ligands with enhanced binding affinity for the sphingosine 1-phosphate receptor 1 using machine learning
Multiple sclerosis (MS) is a debilitating neurological disease affecting nearly one million people in the United States. Sphingosine-1-phosphate receptor 1, or S1PR1, is a protein target for MS…
Thumbnail for Non-Markovian dynamic modeling reveals chromatin fiber destabilization in nucleosome condensates | Poster Board #1513
Non-Markovian dynamic modeling reveals chromatin fiber destabilization in nucleosome condensates | Poster Board #1513
Chromatin's role in governing genome function, encompassing DNA transcription, replication, and repair, is tightly linked to its intricate organizational structure within the nucleus. Despite its significance, deciphering the complexity of chromatin has been challenging…