Deep Neural Variant Classifier for Somatic–Germline Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Next-generation sequencing data presents complex interdependencies that hinder traditional classifiers like support vector machines in variant classification tasks, necessitating advanced classifiers capable of extracting high-level features.
Innovation Solution
A variant classifier utilizing deep neural networks, specifically convolutional neural networks (CNNs) and fully-connected neural networks (FCNNs), processes DNA sequences and correlated metadata features to classify variants as somatic or germline, employing transfer learning and regularization techniques for enhanced performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional classifiers like support vector machines are used for variant classification, then the classification process is simpler and more interpretable, but the ability to handle complex interdependencies and extract high-level features from sequenced data is insufficient
Solution Approach 1:
The patent replaces traditional mechanical classification systems (support vector machines) with deep neural networks that can automatically learn and extract high-level features from complex sequenced data, resolving the contradiction between simplicity and feature extraction capability
Solution Approach 2:
The patent transforms the classification approach by changing from fixed-parameter traditional classifiers to adaptive-parameter deep learning models that can dynamically adjust to capture complex interdependencies in the data
2Measurement precision
If deep neural networks are used for variant classification, then the ability to extract high-level features and handle complex interdependencies is improved, but the computational complexity and resource requirements increase
Solution Approach 1:
The patent segments the complex deep neural network into modular components (convolutional layers, recurrent layers, fully-connected layers) that can be trained and deployed independently, managing computational complexity while maintaining high feature extraction capability
3Reliability
If deep neural networks with multiple layers are used, then the accuracy in distinguishing somatic and germline variants is improved, but the training time and computational resources required increase
Solution Approach 1:
The patent applies transfer learning by pre-training the deep neural network on large-scale genomic data before fine-tuning on specific variant classification tasks, reducing training time while maintaining high accuracy in distinguishing somatic and germline variants
Data Source
AI summary
We introduce a variant classifier that uses trained deep neural networks to predict whether a given variant is somatic or germline. Our model has two deep neural networks: a convolutional neural network (CNN) and a fully-connected neural network (FCNN), and two inputs: a DNA sequence with a variant and a set of metadata features correlated with the variant. The metadata features represent the variant's mutation characteristics, read mapping statistics, and occurrence frequency. The CNN processes the DNA sequence and produces an intermediate convolved feature. A feature sequence is derived by concatenating the metadata features with the intermediate convolved feature. The FCNN processes the feature sequence and produces probabilities for the variant being somatic, germline, or noise. A transfer learning strategy is used to train the model on two mutation datasets. Results establish advantages and superiority of our model over traditional classifiers.