id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
cord-102705-mcit0luk	Gupta, Chitrak	Mind reading of the proteins: Deep-learning to forecast molecular dynamics	2020-07-29		.txt	text/plain	6355	314	51	two types of data sets, the dynamic correlations within which pose significant challenge on existing machine-learning techniques for predicting the real-time nonlinear dynamics of proteins. In this intermediate-dimensional space, where the data distribution is densed highly correlated, we train state-of-the-art time sequence modeling techniques including recurrent neural networks (RNNs) with long short term memory (LSTMs) to predict the future state of the system (Fig. 1 ). We present two new data sets to introduce subtleties in the equilibrium and nonequilibrium molecular dynamics from the perspective of time series forecasting. The assumption is incorrect, but still helps us set a realistic baseline for evaluating the performance of advanced machine learning techniques like LSTMs. Figures 6A,B (ADK) and 8A,B (SMD) show the RMSD distributions of static model for lead time steps 15 and 120, respectively. Protein dynamics was represented as a time-series data and was modeled through a recurrent neural network with LSTM cells in the hidden layer.	./cache/cord-102705-mcit0luk.txt	./txt/cord-102705-mcit0luk.txt