An Efficient Lightweight Hybrid Model with Attention Mechanism for Enhancer Sequence Recognition

Abstract

Enhancers are sequences with short motifs that exhibit high positional variability and free scattering properties. Identification of these noncoding DNA fragments and their strength are extremely important because they play a key role in controlling gene regulation on a cellular basis. The identification of enhancers is more complex than that of other factors in the genome because they are freely scattered, and their location varies widely. In recent years, bioinformatics tools have enabled significant improvement in identifying this biological difficulty. Cell line-specific screening is not possible using these existing computational methods based solely on DNA sequences. DNA segment chromatin accessibility may provide useful information about its potential function in regulation, thereby identifying regulatory elements based on its chromatin accessibility. In chromatin, the entanglement structure allows positions far apart in the sequence to encounter each other, regardless of their proximity to the gene to be acted upon. Thus, identifying enhancers and assessing their strength is difficult and time-consuming. The goal of our work was to overcome these limitations by presenting a convolutional neural network (CNN) with attention-gated recurrent units (AttGRU) based on Deep Learning. It used a CNN and one-hot coding to build models, primarily to identify enhancers and secondarily to classify their strength. To test the performance of the proposed model, parallels were drawn between enhancer-CNNAttGRU and existing state-of-the-art methods to enable comparisons. The proposed model performed the best for predicting stage one and stage two enhancer sequences, as well as their strengths, in a cross-species analysis, achieving best accuracy values of 87.39% and 84.46%, respectively. Overall, the results showed that the proposed model provided comparable results to state-of-the-art models, highlighting its usefulness.

A comprehensive revisit of the machine‐learning tools developed for the identification of enhancers in the human genome

“…In our comprehensive review, approximately 75% of the proposed methods utilized the Liu dataset, some used the FANTOM5 dataset, and one utilized EnhancerAtlas 2.0. Notably, three new methods were published in 2023 using the Liu dataset2 [142–144]. Unfortunately, we are unable to provide a specific reason why researchers prefer the Liu dataset rather than creating their dataset.…”

Enhancing Short-Term Electrical Load Forecasting for Sustainable Energy Management in Low-Carbon Buildings

“…Likewise, in [25,43], a hybrid model was proposed in which two models, a CNN and Gated-Recurrent-Unit (GRU), were integrated. Furthermore, other popular electricity load forecasting models have been created, such as the multivariable Experimental Mode Decomposition network, an expanded CNN in combination with LSTM [44], hybrid artificial neural network models [20,45], Empirical Mode Decomposition (EMD), and Extreme Learning Machine (ELM) models [46]. These approaches overcome the issues of shallow-based models, but they are unreliable for accurate implementation due to low forecasting accuracy.…”

MedicalCLIP: Anomaly-Detection Domain Generalization with Asymmetric Constraints

“…Anomaly detection is widely applied across various sectors including industrial production [1][2][3], finance, autonomous driving [4], and disease diagnosis [5][6][7][8][9][10]. In the medical field, anomaly detection can help reduce misdiagnoses and missed diagnoses caused by human error during manual inspections.…”