Motivation:Enhancers are distal cis-acting regulating regions that play a vital role in gene transcription. However, due to the inherent nature of enhancers being linearly distant from the affected gene in an irregular manner while being spatially close at the same time, systematically predicting enhancers has been a challenging task. Although several computational predictor models through both epigenetic marker analysis and sequence-based analysis have been proposed, they lack generalization capacity across different enhancer datasets and have feature dependency. On the other hand, the recent proliferation of deep learning methods has opened previously unknown avenues of approach for sequence analysis tasks which eliminates feature dependency and achieves greater generalization. Therefore, harnessing the power of deep learning based sequence analysis techniques to develop a more generalized model than the ones developed before to predict enhancer region in a DNA sequence is a topic of interest in bioinformatics.
Results:In this study, we develop the predictor model CHilEnPred that has been trained with the visual representation of the DNA sequences with Hilbert Curve. We report our computational prediction result on FANTOM5 dataset where CHilEnPred achieves an accuracy of 94.97% and AUC of 0.987 on test data. Availability: Our CHilEnPred model can be freely accessed at https://github.com/iatahmid/chilenpred Contact: msrahman@cse.buet.ac.bd
Genome wide association studies (GWAS) attempt to map genotypes to phenotypes in organisms. This is typically performed by genotyping individuals using microarray or by aligning whole genome sequencing reads to a reference genome. Both approaches require knowledge of a reference genome which hinders their application to organisms with no or incomplete reference genomes. This caveat can be removed by using alignment-free association mapping methods based on k-mers from sequencing reads. Here we present an improved implementation of an alignment free association mapping method. The new implementation is faster and includes additional features to make it more flexible than the original implementation. We have tested our implementation on an E. Coli ampicillin resistance dataset and observe improvement in execution time over the original implementation while maintaining accuracy in results. We also demonstrate that the method can be applied to find sex specific sequences.
Deep Neural Network (DNN) classifiers are known to be vulnerable to Trojan or backdoor attacks, where the classifier is manipulated such that it misclassifies any input containing an attacker-determined Trojan trigger. Backdoors compromise a model's integrity, thereby posing a severe threat to the landscape of DNN-based classification. While multiple defenses against such attacks exist for classifiers in the image domain, there have been limited efforts to protect classifiers in the text domain.We present Trojan-Miner (T-Miner) -a defense framework for Trojan attacks on DNN-based text classifiers. T-Miner employs a sequence-to-sequence (seq-2-seq) generative model that probes the suspicious classifier and learns to produce text sequences that are likely to contain the Trojan trigger. T-Miner then analyzes the text produced by the generative model to determine if they contain trigger phrases, and correspondingly, whether the tested classifier has a backdoor. T-Miner requires no access to the training dataset or clean inputs of the suspicious classifier, and instead uses synthetically crafted "nonsensical" text inputs to train the generative model. We extensively evaluate T-Miner on 1100 model instances spanning 3 ubiquitous DNN model architectures, 5 different classification tasks, and a variety of trigger phrases. We show that T-Miner detects Trojan and clean models with a 98.75% overall accuracy, while achieving low false positives on clean models. We also show that T-Miner is robust against a variety of targeted, advanced attacks from an adaptive attacker.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.