DARTS Analysis

Introduction of DARTS Analysis

An important feature of eukaryotic genes that distinguishes them from prokaryotes is that they have introns. The presence of introns makes the expression of eukaryotic genes have to undergo the important step of RNA splicing. A gene in higher eukaryotes contains multiple introns. Through the regulation of RNA splicing, a gene can be transcribed into multiple different transcripts, thereby encoding multiple proteins, increasing the complexity and adaptability of life. Although RNA-seq technology can better quantify the results of gene expression, it requires a higher sequencing depth for differential splicing analysis. Therefore, for general-depth RNA-seq data, the existing calculation methods cannot accurately quantify gene splicing.

In order to improve the accuracy of RNA-seq quantitative differential splicing, a new computing framework DARTS (Deep-learning Augmented RNA-seq analysis of Tran Splicing) was born with the help of machine learning methods. DARTS combines deep learning with Bayesian hypothesis testing for RNA alternative splicing analysis. It leverages public RNA-seq big data to provide a knowledge base of splicing regulation via deep learning, helping researchers better characterize alternative splicing using RNA-seq datasets even with modest coverage. Accurate results of alternative splicing analysis will have far-reaching significance for further understanding of the regulatory mechanism of alternative splicing in different cell states.

Fig 1. The DARTS computational framework for deep learning-augmented RNA-seq analysis of transcript splicing. (Zhang Z, et al. 2019)

DARTS Analysis Principle

DARTS consists of two parts: Bayesian hypothesis testing (BHT) framework and Deep Neural Network Module (DNN):

DARTS BHT, a Bayesian statistical framework to determine the statistical significance of differential splicing events or unchanged splicing events between RNA-seq data of two biological conditions. The DARTS BHT framework was designed to integrate deep learning based prediction as prior and empirical evidence in a specific RNA-seq dataset as likelihood.
Deep neural network model, a core component of the DARTS BHT framework, generates a probability of differential splicing between two biological conditions using exon- and sample-specific predictive features. The DARTS DNN were designed to predict differential splicing of a given exon based on exon-specific cis sequence features and sample-specific trans RBP expression levels in two biological conditions.

Advantages of DARTS Analysis

DARTS DNN not only predicts the results of alternative splicing by cis sequence features, but also integrates the expression level of RNA binding proteins in the sample into the prediction of the results of RNA alternative splicing, increasing the dimension of the prediction parameters.
Through DNN's deep learning of a large number of RNA-seq results in the database, high-precision predicted values can be obtained as the Bayesian prior probability in BHT. Then combined with specific RNA-seq data analysis to obtain more accurate differential splicing results.
Even in low-throughput RNA-seq libraries, enhanced analysis by using DNN predicted values can achieve higher accuracy than traditional methods, and this improvement is more obvious in lower-throughput libraries. Even in a high-throughput RNA-seq library, using DNN prediction can still find alternative splicing changes in low-expressed genes.

Application Filed

Research on the occurrence and development of various diseases.
Drug discovery and research.
Analysis of alternative splicing in RNA-seq in animals and plants.

CD ComputaBio provides RNA-seq analysis of alternative splicing based on DARTS. With the help of artificial intelligence methods, DARTS provides a better means of alternative splicing analysis for data with low sequencing depth, and expands the sensitivity and accuracy of traditional RNA-seq alternative splicing analysis. In addition to DARTS, we also provide other types of software for customers to analyze alternative splicing data. for data analysis, you only need to provide the original data, and we will conduct a complete data analysis based on your research. If you have any questions, please feel free to contact us, we have a professional technical support team to provide you with services.

References

Zhang Z, et al. Deep-learning augmented RNA-seq analysis of transcript splicing. [J]. Nat Methods. 2019 Apr;16(4):307-310.
Baralle F E, Giudice J. Alternative splicing as a regulator of development and tissue identity[J]. Nature Reviews Molecular Cell Biology, 2017, 18(7).

DARTS Analysis

DARTS Analysis Principle

Advantages of DARTS Analysis

Application Filed

References

Our Featured Solutions