Deep Learning Using Havrda-Charvat Entropy for Classification of Pulmonary Optical Endomicroscopy

1) ObjectivePulmonary optical endomicroscopy (POE) is an imaging technology in real time. It allows to examine pulmonary alveoli at a microscopic level. Acquired in clinical settings, a POE image sequence can have as much as 25% of the sequence being uninformative frames (i.e. pure-noise and motion artifacts). For future data analysis, these uninformative frames must be first removed from the sequence. Therefore, the objective of our work is to develop an automatic detection method of uninformative images in endomicroscopy images.2) Material and methodsWe propose to take the detection problem as a classification one. Considering advantages of deep learning methods, a classifier based on CNN (Convolutional Neural Network) is designed with a new loss function based on Havrda-Charvat entropy which is a parametrical generalization of the Shannon entropy. We propose to use this formula to get a better hold on all sorts of data since it provides a model more stable than the Shannon entropy.3) ResultsOur method is tested on one POE dataset including 3895 distinct images and is showing better results than using Shannon entropy and behaves better with regard to the problem of overfitting. We obtain 70% of accuracy with Shannon entropy versus 77 to 79% with Havrda-Charvat.4) ConclusionWe can conclude that Havrda-Charvat entropy is better suited for restricted and or noisy datasets due to its generalized nature. It is also more suitable for classification in endomicroscopy datasets.     

基于卷积神经网络的大肠杆菌启动子预测

目的 基于位点特异性打分矩阵（position-specific scoring matrices,PSSM）的预测模型已经取得了良好的效果,基于PSSM的各种优化方法也在不断发展,但准确率相对较低,为了进一步提高预测准确率,本文基于卷积神经网络（convolutional neural networks,CNN）算法做了进一步研究。方法 采用PSSM将启动子序列处理成数值矩阵,通过CNN算法进行分类。大肠杆菌K-12（Escherichia coli K-12,E.coli K-12,下文简称大肠杆菌）的Sigma38、Sigma54和Sigma70 3种启动子序列被作为正集,编码（Coding）区和非编码（Non-coding）区的序列为负集。结果 在预测大肠杆菌启动子的二分类中,准确率达到99%,启动子预测的成功率接近100%;在对Sigma38、Sigma54、Sigma70 3种启动子的三分类中,预测准确率为98%,并且针对每一种序列的预测准确率均可以达到98%以上。最后,本文以Sigma38、Sigma54、Sigma70 3种启动子分别和Coding区或者Non-coding区序列做四分类,预测得到的准确性为0.98,对3种Sigma启动子均衡样本的十交叉检验预测精度均可以达到0.95以上,海明距离为0.016,Kappa系数为0.97。结论 相较于支持向量机（support vector machine,SVM）等其他分类算法,CNN分类算法更具优势,并且基于CNN的分类优势,编码方式亦可以得到简化。     

基于CNN与LSTM模型的蛋白质二级结构预测

蛋白质结构的预测在理解蛋白质结构组成和蛋白质的生物学功能有重要意义,而蛋白质二级结构预测是蛋白质结构预测的重要环节。当PSSM位置特异性进化矩阵被广泛应用于将蛋白质初级结构序列编码作为输入样本后,每个残基可以被表示成二维空间的数据平面,由此文中尝试利用卷积神经网络对其进行训练。文中还设计了另一种卷积神经网络,利用长短记忆网络感知了CNN最后卷积特征面的横向特征和纵向特征后连同卷积神经网络的全连接共同完成分类,最后用ensemble方法对两类卷积神经网络模型进行了整合,最终ensemble方法中包含两类卷积神经网络的六个模型,在CB513蛋白质数据集测得的Q3结果为77.2。     

Genetics and mechanisms of hepatic cystogenesis

Polycystic liver disease (PLD) is a heterogeneous genetic condition. PKD1 and PKD2 germline mutations are found in patients with autosomal dominant polycystic kidney disease (ADPKD). Autosomal dominant polycystic liver disease (ADPLD) is associated with germline mutations in PRKCSH, SEC63, LRP5, and recently ALG8 and SEC61. GANAB mutations are found in both patient groups. Loss of heterozygosity of PLD-genes in cyst epithelium contributes to the development of hepatic cysts. A genetic interaction network is implied in hepatic cystogenesis that connects the endoplasmic glycoprotein control mechanisms and polycystin expression and localization. Wnt signalling could be the major downstream signalling pathway that results in hepatic cyst growth. PLD in ADPLD and ADPKD probably results from changes in one common final pathway that initiates cyst growth. This article is part of a Special Issue entitled: Cholangiocytes in Health and Diseaseedited by Jesus Banales, Marco Marzioni, Nicholas LaRusso and Peter Jansen.     

Identification of Indian butterflies using Deep Convolutional Neural Network

The conventional butterfly identification method is based on their different morphological characters namely wing-venation, color, shape, patterns and through the dissection studies and molecular techniques which are tedious, expensive and highly time-consuming. To overcome the above aforesaid challenges, a new butterfly identification system using butterfly images has been designed to instantly identify the butterfly with high accuracy. In this study, we construct a new butterfly dataset with 34,024 butterfly images belonging to 315 species from India. We propose and prove the effectiveness of new data augmentation techniques on our dataset. To identify butterflies using photographic images, we built eleven new Deep Convolutional Neural Network (DCNN) butterfly classifier models using eleven pre-trained architectures namely ResNet-18, ResNet-34, ResNet-50, ResNet-121, ResNet-152, Alex-Net, DenseNet-121, DenseNet-161, VGG-16, VGG-19 and SqueezeNet-v1.1. The different model's classification results were compared and the proposed technique achieved a maximum top-1 accuracy(94.44%), top-3 accuracy(98.46%) and top-5 accuracy(99.09%) using ResNet-152 model, followed by DenseNet-161 model achieved the top-1 accuracy(94.31%), top-3 accuracy (98.07%) and top-5 accuracy (98.66%). The results suggest that models can be assertively used to identify butterflies in India.     

New Intraclass Helitrons Classification Using DNA-Image Sequences and Machine Learning Approaches

Helitrons, eukaryotic transposable elements (TEs) transposed by rolling-circle mechanism, have been found in various species with highly variable copy numbers and sometimes with a large portion of their genomes. The impact of helitrons sequences in the genome is to frequently capture host genes during their transposition. Since their discovery, 18 years ago, by computational analysis of whole genome sequences of Arabidopsis thaliana plant and Caenorhabditis elegans (C. elegans) nematode, the identification and classification of these mobile genetic elements remain a challenge due to the fact that the wide majority of their families are non-autonomous. In C. elegans genome, DNA helitrons sequences possess great variability in terms of length that varies between 11 and 8965 base pairs (bps) from one sequence to another. In this work, we develop a new method to predict helitrons DNA-sequences, which is particularly based on Frequency Chaos Game Representation (FCGR) DNA-images. Thus, we introduce an automatic system in order to classify helitrons families in C. elegans genome, based on a combination between machine learning approaches and features extracted from DNA-sequences. Consequently, the new set of helitrons features (the FCGR images and K-mers) are extracted from DNA sequences. These helitrons features consist of the frequency apparition number of K nucleotides pairs (Tandem Repeat) in the DNA sequences. Indeed, three different classifiers are used for the classification of all existing helitrons families. The results have shown potential global score equal to 72.7% due to FCGR images which constitute helitrons features and the pre-trained neural network as a classifier. The two other classifiers demonstrate that their efficiency reaches 68.7% for Support Vector Machine (SVM) and 91.45% for Random Forest (RF) algorithms using the K-mers features corresponding to the genomic sequences.     

BEguider:一个单碱基编辑器sgRNA设计与编辑效率预测工具

单碱基编辑器是实用且高效的基因编辑工具,其编辑效率与单向导RNA(single guide RNA, sgRNA)序列的设计密切相关。目前单碱基编辑器sgRNA序列的设计缺少特定的法则,主要依靠经验和大量尝试完成。本研究基于卷积神经网络,开发了一个单碱基编辑器sgRNA序列设计工具BEguider。BEguider利用TensorFlow 2深度学习框架建立编辑效率预测模型,能够在人基因组范围内针对NGG PAM序列依赖的单碱基编辑器ABE7.10-NGG和BE4-NGG批量设计sgRNA序列,预测编辑效率。此外,通过整合Cas-OFFinder, BEguider能够提供对sgRNA脱靶情况的评估。利用BEguider设计sgRNA序列,有助于研究人员提高实验效率,节约实验成本。     

A novel multi-head CNN design to identify plant diseases using the fusion of RGB images

Plant diseases and insect pests cause a significant threat to agricultural production. Early detection and diagnosis of these diseases are critical and can reduce economic losses. The recent development of deep learning (DL) benefits various fields, such as image processing, remote sensing, medical diagnosis, and agriculture. This work proposed a novel approach based on DL for plant disease detection by fusing RGB and segmented images. A multi-headed DenseNet-based architecture was developed, considering two images as input. We evaluated the model on a public dataset, PlantVillage, consisting of 54183 images with 38 classes. The fivefold cross-validation technique achieved an average accuracy, recall, precision, and f1-score of 98.17%, 98.17%, 98.16%, and 98.12%, respectively. The proposed approach can distinguish various plant diseases with different characteristics by image fusion. The high success rate with low standard deviation proves the robustness of the model, and the model can be integrated into plant disease detection and early warning system.     

Integrative gene network analysis provides novel regulatory relationships, genetic contributions and susceptible targets in autism spectrum disorders

Autism spectrum disorders (ASDs) are a group of diseases exhibiting impairment in social drive, communication/language skills and stereotyped behaviors. Though an increased number of candidate genes and molecular interactions have been identified by various approaches, the pathogenesis remains elusive. Based on clinical observations, data from accessible GWAS and expression datasets we identified ASDs gene candidates. Integrative gene network and a novel CNV-centric Node Network (CNN) analysis method highlighted ASDs-associated key elements and biological processes. Functional analysis identified neurological functions including synaptic cholinergic receptor (CHRNA) families, dopamine receptor (DRD2), and correlations between social behavior and oxytocin related pathways. CNN analysis of genome-wide genetic and expression data identified inheritance-related clusters related to PTEN/TSC1/FMR1 and mTor/PI3K regulation. Integrative analysis identified potential regulators of networks, specifically TNF and beta-estradiol, suggesting a potential central role in ASDs. Our data provide information on potential disease mechanisms, and key regulators that may generate novel postulations, and diagnostic molecular biomarkers.