2021/09/13 by Michele Monti, Jonathan Fiorentino, Monti, Michele +7
Biochemistry, Genetics and Molecular Biology · Computer Science · Physics and Astronomy · #Artificial intelligence #Artificial neural network #Biological Physics (physics.bio-ph) #Biology #Computer science #Data Analysis #Evolutionary Algorithms and Applications #FOS: Biological sciences #FOS: Computer and information sciences #FOS: Physical sciences #Gene #Gene Regulatory Network Analysis #Gene expression #Gene expression and cancer classification #Gene regulatory network #Genetics #Graph #Inference #Interpretability #Machine Learning (cs.LG) #Machine learning #Molecular Networks (q-bio.MN) #Neural Networks and Applications #Quantitative Methods (q-bio.QM) #Recurrent neural network #Statistics and Probability (physics.data-an) #Theoretical computer science #Time series #cs.LG #physics.bio-ph #physics.data-an #q-bio.MN #q-bio.QM
paper · pdf · doi:10.48550/arxiv.2109.05849
17 pages, 6 figures
arxiv created 2021/09/13 · openalex publication_date 2021/09/13 · arxiv updated 2021/09/14 · openalex created_date 2022/07/25 · openalex updated_date 2026/08/05
Methods for time series prediction and classification of gene regulatory\nnetworks (GRNs) from gene expression data have been treated separately so far.\nThe recent emergence of attention-based recurrent neural networks (RNN) models\nboosted the interpretability of RNN parameters, making them appealing for the\nunderstanding of gene interactions. In this work, we generated synthetic time\nseries gene expression data from a range of archetypal GRNs and we relied on a\ndual attention RNN to predict the gene temporal dynamics. We show that the\nprediction is extremely accurate for GRNs with different architectures. Next,\nwe focused on the attention mechanism of the RNN and, using tools from graph\ntheory, we found that its graph properties allow to hierarchically distinguish\ndifferent architectures of the GRN. We show that the GRNs respond differently\nto the addition of noise in the prediction by the RNN and we relate the noise\nresponse to the analysis of the attention mechanism. In conclusion, this work\nprovides a a way to understand and exploit the attention mechanism of RNN and\nit paves the way to RNN-based methods for time series prediction and inference\nof GRNs from gene expression data.\n