Current Bioinformatics - Current Issue

The coronavirus disease COVID-19, caused by the SARS-CoV-2 virus, was a global pandemic that happened in March of 2020. The virus was mutated into several widely-spread strains such as Alpha, Beta, Gamma, Delta, and Omicron, and is continuing its unpredictable mutation.

Multi-Epitope Vaccine (MEV) is one type of recombinant vaccine with its sequence containing multiple epitopes and is considered as an effective way to fight against the infectious disease. Previous in silico approaches to MEV construction have been constrained by their inability to predict molecular conformation structures accurately, consequently leading to inaccurate property evaluations. In this work, we designed a novel MEV for the future prevention of COVID-19 or similar diseases. We set strict thresholds to screen for epitope candidates in order to construct highly effective MEV and use the latest ColabFold (a modified version of AlphaFold2) to predict accurate tertiary structures of the MEV.

We especially studied epitopes from the main proteins of SARS-CoV-2 (i.e., the envelope, membrane, nucleo-, and spike proteins) that can provoke immunity response of B-cells, helper T-cells (Th), and cytotoxic T-cells (CTL), then we combined them through amino acid linkers to construct the MEV. We evaluated the vaccine in terms of its physicochemical properties, population coverage, safety for use, secondary and tertiary structure, docking immunity response, and immu nity response eliciting capability.

These in silico assessments demonstrate that our proposed vaccine can elicit effective immune responses and it is safe to use with a high population coverage.

One of the main causes of cancer-related mortality in women is breast cancer (BC). There were four molecular subtypes of this malignancy, and adjuvant therapy efficacy differed based on these subtypes. Gene expression profiles provide valuable information that is helpful for patients whose prognosis is not clear from clinical markers and immunohistochemistry.

In this study, we aim to predict molecular types of BC using a gene expression dataset of patients with BC and normal samples using six well-known ensemble machine-learning techniques.

Two microarray datasets were downloaded; (GSE45827) and (GSE140494) from the Gene Expression Omnibus (GEO) database. These datasets comprise 21 samples of normal tissues that were part of a cohort analysis of primary invasive breast cancer (57 basal, 36 HER2, 56 Luminal A, and 66 Luminal B). Namely, we used AdaBoost, Random Forest (RF), Artificial Neural Network (ANN), Naïve Bayes (NB), Classification and Regression Tree (CART), and Linear Discriminant Analysis (LDA) classifiers.

The results of the data analysis show that the RF and NB classifiers outperform the other models in the prediction of the BC subtype. The RF shows superior performance with an accuracy range between 0.89 and 1.0 in contrast to its competitor NB, which has an average accuracy of 0.91. Our approach perfectly discriminates un-affected cases (normal) from the carcinoma. In this case, the RF provides perfect prediction with zero errors. Additionally, we used PCA, DHWT low-frequency, and DHWT high-frequency to perform a dimensional reduction for the numerous gene expression values. Consequently, the LDA achieves up to 95% improvement in performance through data reduction. Moreover, feature selection allowed for the best performance, which is recorded by the RF with classification accuracy 98%.

Overall, we provide a successful framework that leads to shorter computation times and smaller ML models, especially where memory and time restrictions are crucial.

With the rapid evolution of single-cell RNA sequencing technology, the study of cellular heterogeneity in complex tissues has reached an unprecedented resolution. One critical task of the technology is cell-type annotation. However, challenges persist, particularly in annotating novel cell types.

Current methods rely heavily on well-annotated reference data, using correlation comparisons to determine cell types. However, identifying novel cells remains unstable due to the inherent complexity and heterogeneity of scRNA-seq data and cell types. To address this problem, we propose scADCA, a method based on anomaly detection, for identifying novel cell types and annotating the entire dataset.

The convolutional modules and fully connected networks are integrated into an autoencoder, and the reference dataset is trained to obtain the reconstruction errors. The threshold based on these errors can distinguish between novel and known cells in the query dataset. After novel cells are identified, a multinomial logistic regression model fully annotates the dataset.

Using a simulation dataset, three real scRNA-seq pancreatic datasets, and a real scRNA-seq lung cancer cell line dataset, we compare scADCA with six other cell-type annotation methods, demonstrating competitive performance in terms of distinguished accuracy, full accuracy, F1 -score, and confusion matrix.

In conclusion, the scADCA method can be further improved and expanded to achieve better performance and application effects in cell type annotation, which is helpful to improve the accuracy and reliability of cytology research and promote the development of single-cell omics.

Long non-coding RNAs (lncRNAs) are a category of more extended RNA strands that lack protein-coding abilities. Although they are not involved in the translation of proteins, studies have shown that they play essential regulatory functions in cells, regulating gene expression and cell biological processes. However, it is both costly and inefficient to determine the associations between lncRNAs and diseases through biological experiments. Therefore, there is an urgent need to develop convenient and fast computational methods to predict lncRNA-disease associations (LDAs) more efficiently.

Predicting disease-associated lncRNAs can help explore the mechanisms of action of lncRNAs in diseases, and this is crucial for early intervention and treatment of diseases.

In this paper, we propose an enhanced heterogeneous graph representation method for predicting LDAs, named GCGALDA. The GCGALDA first obtains the topological structure features of nodes by a biased random walk. Based on this, the neighboring nodes of a node are weighted using the attention mechanism to further mine the semantic association relationships between nodes in the graph data. Then, a graph convolution network (GCN) is used to transfer the neighborhood features of the node to the central node and combine them with the node's features so that the final node representation contains not only structural information but also semantic association information. Finally, the association score between lncRNA and disease is obtained by multilayer perceptron (MLP).

As evidenced by the experimental findings, the GCGALDA outperforms other advanced models in terms of prediction accuracy on openly accessible databases. In addition, case studies on several human diseases further confirm the predictive ability of the GCGALDA.

In conclusion, the proposed GCGALDA model extracts multi-perspective features, such as topology, semantic association, and node attributes, obtains high-quality heterogeneous graph node representations, and effectively improves the performance of the LDA prediction model.

Hepatocellular Carcinoma (HCC) is a major disease that seriously threatens human health. Early screening can significantly improve the five-year survival rate of HCC patients. Cell-free DNA (cfDNA), as a potential carrier of cancer signals in body fluids, can be used for early cancer detection. However, current early HCC detection methods based on cfDNA sequencing require deep sequencing data, limiting their application and usage in routine disease screening. We proposed a foundational DNA language model, called CLHCC, for analyzing DNA sequences and methylation patterns to detect HCC at low sequencing depths.

CLHCC randomly selected 1500 DNA fragments from HCC-specific differentially methylated regions identified by cd-score. The model then performed a one-hot encoding strategy on these DNA fragments and input the data into a CNN combined with an LSTM neural network for classification.

We tested CLHCC on 2139 target-BS data samples, achieving an accuracy of 84.59% (precision: 83.44%, recall: 81%) under 10-fold cross-validations. This performance is better than DNA language models built using CNN or LSTM alone.

Our study applies deep learning to analyze DNA sequences in specific methylation regions without the need for complex alignment processes. This provides new theoretical and practical guidance for clinical applications and holds promise for non-invasive early HCC screening via cfDNA.