Using deep learning to model the hierarchical structure and function of a cell

By Jianzhu Ma, Michael Ku Yu, Samson H. Fong, Keiichiro Ono, Eric Sage, Barry Demchak, Roded Sharan, Trey Ideker. Nature Methods, 2018

391 citations recorded by the source corpus.

Although artificial neural networks are powerful classifiers, their internal structures are hard to interpret. In the life sciences, extensive knowledge of cell biology provides an opportunity to design visible neural networks (VNNs) that couple the model's inner workings to those of real systems. Here we develop DCell, a VNN embedded in the hierarchical structure of 2,526 subsystems comprising a eukaryotic cell (http://d-cell.ucsd.edu/). Trained on several million genotypes, DCell simulates cellular growth nearly as accurately as laboratory observations. During simulation, genotypes induce patterns of subsystem activities, enabling in silico investigations of the molecular mechanisms underlying genotype–phenotype associations. These mechanisms can be validated, and many are unexpected; some are governed by Boolean logic. Cumulatively, 80% of the importance for growth prediction is captured by 484 subsystems (21%), reflecting the emergence of a complex phenotype. DCell provides a foundation for decoding the genetics of disease, drug resistance and synthetic life.

Subfields: Artificial Intelligence, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/0a063d1ec630a7cecaad253c34f190cc2db2241e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

Citations in the corpus

Machine Learning for Integrating Data in Biology and Medicine: Principles, Practice, and Opportunities (2018, methodology): “In a similar way, [242] used a neural network and integrated into the neural model prior biological knowledge given by the Gene Ontology [203].”
Deep-learning seismology (2022, neutral): “In this approach, the network architecture is designed to provide insight by promoting or restricting specific attributes of the data being modeled ( 146 , 147 ).”
A Survey on Deep Learning (2018, methodology): “Ma et al. explain neural networks using cell biology from the molecular scale up [104].”
A white-box machine learning approach for revealing antibiotic mechanisms of action (2019, neutral): “Investigation into these other physiological systems will require new and different modeling approaches (Carrera and Covert, 2015; Ma et al., 2018; Oberhardt et al., 2013; Yang et al., 2018), curated knowledge bases (Karr et al.”
The BioGRID interaction database: 2019 update (2018, background): “Biological interaction networks, as aggregated from a plethora of individual protein or genetic interactions, as well as interactions of RNA, DNA, membranes, carbohydrates and small molecule metabolites, serve as a framework for understanding gene–phenotype relationships and the mechanistic basis fo”
The BioGRID database: A comprehensive biomedical resource of curated protein, genetic, and chemical interactions (2020, methodology): “For this reason, BioGRID interaction data have frequently been used as a gold standard in computational studies, either as training sets for machine learning or for verification of predicted results.(3,7,8) BioGRID houses interactions from all major model organisms (Figure 1b, Table 1).”
Deep learning in bioinformatics: introduction, application, and perspective in big data era (2019, background): “Biomolecule interaction prediction and systems biology Microarray gene expression, PPI, gene-disease interaction, diseasedisease similarity network, diseasevariant network 1D data, 2D data, structured data, graph data CNN, GCN [95, 201, 203, 165, 71, 191, 67, 88]”
Obtaining genetics insights from deep learning via explainable artificial intelligence (2022, background): “In a similar spirit to DCell and P-NET, others have created neural networks in which units and parameters have a biophysical interpretation 80,81 .”
Scaling Vision Transformers to Gigapixel Images via Hierarchical Self-Supervised Learning (2022, background): “Broader Impacts: Many problems in biology and medicine have hierarchical-like relationships [34, 48, 55].”
ARTIFICIAL INTELLIGENCE AND PERSONALIZED MEDICINE (2019, background): “[ 13] Finally, web sites with large databases like DeepChem (https://deepchem.io/about.html), which leverage AI in chemistry settings, can be used specifically to identify properties of drugs and compounds.”
Measuring biological age using omics data (2022, background): “Nevertheless, there is progress in designing interpretable neural networks by explicitly encoding specific biological pathways into their structure 202 or back-propagating information to input features 203 ; these exciting approaches may better capture the complexity of ageing. composite rate of…”
Systems Biology of Cancer Metastasis (2019, methodology): “Dcell uses deep neural networks to model the growth phenotype of gene deletion, while explaining the prediction in terms of a chain of causal changes in different modules such as subcellular structures and pathways(Ma et al., 2018).”
MOLI: multi-omics late integration with deep neural networks for drug response prediction (2019, neutral): “The datasets used were from different resources were not in the same format and required substantial pre-processing and standardization (see the Supplementary Material). For example, different studies used different pipelines to detect CNA and reported different estimates of copy number which could ”