This paper introduces a novel finite Zak transform (FZT)-aided framework for constructing multiple zero-correlation zone (ZCZ) sequence sets with optimal correlation properties. Specifically, each sequence is perfect with zero auto-correlation sidelobes, each ZCZ sequence set meets the Tang-Fan-Matsufuji bound with equality, and the maximum inter-set cross-correlation of multiple sequence sets meets the Sarwate bound with equality. Our study shows that these sequences can be sparsely expressed in the Zak domain through properly selected index and phase matrices. Particularly, it is found that the maximum inter-set cross-correlation beats the Sarwate bound if every index matrix is a circular Florentine array. Several construction methods of multiple ZCZ sequence sets are proposed, demonstrating both the optimality and high flexibility. Additionally, it is shown that excellent synchronization performance can be achieved by the proposed sequences in orthogonal-time-frequency-space (OTFS) systems.

The identification of protein-drug interaction plays an important role in the early stage of drug discovery. Recently, deep learning has been introduced into this area and gained excellent results, which mainly benefits from effective representations of protein and drug. However, the sparseness of raw input data builds the fundamental barrier for protein and drug representation learning. In this paper, we propose an end-to-end model representing better the characteristics of protein and drug to estimate binding affinity between protein and drug. By incorporating protein sequence, protein contact map and drug SMILES, our model achieves better results on PDBbind dataset as compared to 3D structure based classical deep model Pafnucy. Moreover, we explore the effects of different parts of our model with ablation study and find that by combining all the modules of our model we can get the best predictive results.

This paper presents a novel method for imputing missing data of multivariate time series by adapting the Long Short Term-Memory(LSTM) and Denoising Autoencoder(DAE). Missing data are ubiquitous in many domains; proper imputation methods can improve performance on many tasks. Our method focus on multivariate time series, applying bidirectional LSTM to learn temporal information and DAE to learn correlation between variables, and we combine these two models by using LSTM as the encoder component of DAE. Several real-world datasets, including electroencephalogram(EEG), electromyogram(EMG) and electronic health records(EHRs), are extracted to test the performance of our method. Through simulation studies, we compare the proposed recurrent denoising autoencoder with several baseline imputation methods and demonstrate its effectiveness in both missing data estimation and label prediction after imputation.