Deep convolutional neural networks for annotating gene expression patterns in the mouse brain

Abstract

BackgroundProfiling gene expression in brain structures at various spatial and temporal scales is essential to understanding how genes regulate the development of brain structures. The Allen Developing Mouse Brain Atlas provides high-resolution 3-D in situ hybridization (ISH) gene expression patterns in multiple developing stages of the mouse brain. Currently, the ISH images are annotated with anatomical terms manually. In this paper, we propose a computational approach to annotate gene expression pattern images in the mouse brain at various structural levels over the course of development.ResultsWe applied deep convolutional neural network that was trained on a large set of natural images to extract features from the ISH images of developing mouse brain. As a baseline representation, we applied invariant image feature descriptors to capture local statistics from ISH images and used the bag-of-words approach to build image-level representations. Both types of features from multiple ISH image sections of the entire brain were then combined to build 3-D, brain-wide gene expression representations. We employed regularized learning methods for discriminating gene expression patterns in different brain structures. Results show that our approach of using convolutional model as feature extractors achieved superior performance in annotating gene expression patterns at multiple levels of brain structures throughout four developing ages. Overall, we achieved average AUC of 0.894 ± 0.014, as compared with 0.820 ± 0.046 yielded by the bag-of-words approach.ConclusionsDeep convolutional neural network model trained on natural image sets and applied to gene expression pattern annotation tasks yielded superior performance, demonstrating its transfer learning property is applicable to such biological image sets.

Supervised and Unsupervised End-to-End Deep Learning for Gene Ontology Classification of Neural In Situ Hybridization Images

“…What is also remarkable is that we were able not only to improve the accuracy of the classification, but to reduce the size of the compact representative vector by more than 10% relatively to [ 12 ], and by more than 80% with respect to the previous deep learning approach in [ 38 ], i.e., to obtain a vector length of 1800 from vector lengths of 2004 and 10,521, respectively.…”

“…Zeng et al [ 38 ] present such a transfer learning approach; in order to obtain a CNN for annotating, eventually, gene expression patterns, they first train a model from OverFeat [ 39 ] on natural images (during the pre-training phase), and then employ this pre-trained network for extracting features of ISH images. This rationale stems from recent studies of using ImageNet data [ 25 ] (an image dataset with thousands of categories and millions of labeled natural images) in training a CNN model, followed by feature extraction (by the model) from other datasets for obtaining overall promising performance.…”

DeNeRD: high-throughput detection of neurons for brain-wide analysis with deep learning

“…Deep Learning (DL) methods are gaining increasing attention in the biological and health sciences. These approaches are used in order to provide fast and comprehensive screening for a change in biological samples over time or in healthy versus potentially diseased human data 12,13 , but so far their use in mouse brain samples is sparse 14–16 . The advantage of AI-based methods over traditional computer vision-based techniques in object detection tasks is their power to capture the variance in an object’s structures without updating the set of hyperparameters for every given data sample.…”

Use of deep neural network ensembles to identify embryonic-fetal transition markers: repression of COX7A1 in embryonic and cancer cells

“…To date, deep learning has been utilized in a variety of biological applications [ 9 ], from prediction of alternate splicing code [ 10 ] to the analysis of protein secondary structure [ 11 ], drug-induced hepatotoxicity [ 12 ], and long non-coding RNAs [ 13 ]. The number of potential applications are, however, more diverse, from basic classification to prediction [ 14 – 16 ], modeling [ 14 ], image processing [ 15 ], and even text mining. Moreover, the complex, noisy, high-dimensional, multi-platform data collated in many biological databases are well suited to deep learning.…”