BiLSTM-VHP: BiLSTM-Powered Network for Viral Host Prediction

Azher Ahmed Efat, Farzana Islam, Annajiat Alim Rasel, Munima Haque · 2025

Read BiLSTM-VHP: BiLSTM-Powered Network for Viral Host Prediction by Azher Ahmed Efat, Farzana Islam, Annajiat Alim Rasel, Munima Haque (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-11345.txt.

Introduction

Azher Ahmed Efat and 3 co-authors predict which animal host a virus came from, directly from its genetic text, in this 2025 paper. Viruses that jump from animals to humans have repeatedly upended public life. Naming the likely host early can help stop spread, but laboratory identification is slow.

Their model, BiLSTM-VHP, reads nucleotide sequences with a bidirectional long short-term memory network, a design suited to patterns in long strings. It was trained and tested on three curated data sets covering orthohantavirus, rabies lyssavirus, and rotavirus A, with tens of thousands of sequences across many host classes. Reported accuracy ran high for rotavirus and orthohantavirus and lower for rabies, beating prior studies on the same task. The data and code are public.

Catalog record

CREATORAzher Ahmed Efat, Farzana Islam, Annajiat Alim Rasel, Munima Haque
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-11345

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

BiLSTM-VHP: BiLSTM-Powered Network for Viral Host Prediction Azher Ahmed Efat, Farzana Islam, Annajiat Alim Rasel, Munima Haque Abstract Recorded history shows the long coexistence of humans and animals, suggesting it began much earlier. Despite some beneficial interdependence, many animals carry viral diseases that can spread to humans. These diseases are known as zoonotic diseases. Recent outbreaks of SARS-CoV-2, Monkeypox and swine flu viruses have shown how these viruses can disrupt human life and cause death. Fast and accurate predictions of the host from which the virus spreads can help prevent these diseases from spreading. This work presents BiLSTM-VHP, a lightweight bidirectional long short-term memory (LSTM)-based architecture that can predict the host from the nucleotide sequence of orthohantavirus, rabies lyssavirus, and rotavirus A with high accuracy. The proposed model works with nucleotide sequences of 400 bases in length and achieved a prediction accuracy of 89.62% for orthohantavirus, 96.58% for rotavirus A, and 77.22% for rabies lyssavirus outperforming previous studies. Moreover, performance of the model is assessed using the confusion matrix, F-1 score, precision, recall, microaverage AUC. In addition, we introduce three curated datasets of orthohantavirus, rotavirus A, and rabies lyssavirus containing 8,575, 95,197, and 22,052 nucleotide sequences divided into 9, 12, and 29 host classes, respectively. The codes and dataset are available at https://doi.org/10.17605/OSF.IO/ANFKR Note: This preprint has not undergone peer review or any post-submission improvements or corrections. The peer-reviewed Version of Record of this contribution is published in International Conference on Advances in Distributed Computing and Machine Learning , and is available online at https://doi.org/10.1007/978-981-96-6721-5_10 1 Introduction Animals play various roles in human society and are an integral part of human history. From being pets to being livestock, those may stay in direct contact with humans in different ways. Alien animals have been identified as zoonotic hosts of one or more zoonoses in studies [ 1 ] and they can transmit those viruses to humans once they come into contact. With deforestation, humans and domestic animals are more likely to be exposed to wild animals. Humans are also in risk of infectious diseases including coming into the contact of rabies lyssavirus, ebola virus, hantavirus due to their adventerous nature [ 2 ] . Due to their evolvability, transmissibility, and lack of therapeutic options, zoonotic viruses pose a great threat [ 3 ] and can occasionally cause pandemics and havoc. Furthermore, emerging diseases are likely to be associated with zoonotic pathogens [ 4 ] , and some changes in human environments may favor these pathogens. In the past, an enormous number of people died due to outbreaks of coronavirus, swine flu, and other viruses. To address an outbreak, it is crucial to find the origin host and isolate it as soon as possible. Continuous research is ongoing in predicting the hosts of viruses. In recent studies, various machine learning algorithms have been utilized to identify influenza A virus origin hosts [ 5 ] . Various representations of the genome has been used to predict the host taxonomic information [ 6 ] [ 7 ] . Additionally, a random forest model can identify the host of coronaviruses based on the whole genome sequence and spike protein sequence [ 8 ] . Other studies have shown the use of SVM and the Mahalanobis distance to predict the host of coronaviruses [ 9 ] . In addition, the hosts of rabies and influenza A viruses were also predicted using SVM [ 10 ] . ML models have also been used to predict GBS host [ 11 ] . However, traditional machine learning algorithms suffer from imbalanced host classes and a lack of annotated virus information. To anticipate the hosts of newly discovered viruses, data scarcity and class imbalance were addressed using transfer learning and ensemble learning [ 12 ] . Additionally, deep neural networks and ensemble classifier were used to identify the host species for viruses [ 13 ] [ 14 ] [ 15 ] . In addition, studies such as VIDHOP provide DNN-based architectures that can predict viral hosts for specific viruses and can also be trained for new viruses [ 16 ] . Although much work has been done on predicting the hosts of novel viruses including SARS-CoV-2 and other common zoonotic viruses such as influenza A virus, very little work has focused on other zoonotic viruses such as orthohantavirus, rabies lyssavirus, and the rotavirus A. HFRS caused by hantavirus is appearing as a serious threat to public health [ 17 ] . Additionally, countless children loss their lives worldwide because of rotavirus each year [ 18 ] . Viral host prediction task has its own significances for many reasons. By knowing which animals are likely to be hosts, health authorities can monitor these species for signs of new or emerging infections. Early detection of viruses in animal hosts can act as a warning for possible human outbreaks, enabling preemptive actions to prevent spillover events. Furthermore, understanding the range of hosts a virus can helps scientists anticipate mutations and design vaccines that offer broader protection. The contributions of this research are as follows: • The introduction of three curated datasets containing nucleotide sequences for rotavirus A, rabies lyssavirus, and orthohantavirus, along with their respective hosts. • A lightweight bidirectional long short-term memory (LSTM)-based model to predict the origin host of rotavirus A, orthohantavirus, and rabies lyssavirus. • To the best of our knowledge, this is the first work on host prediction for orthohantavirus from nucleotide sequences. • Our proposed models outperform those of previous studies in predicting the origin host from nucleotide sequences for rotavirus A as well as for rabies lyssavirus and trained on more nucleotide sequences including latest sequences compared to those of previous studies. 2 Methodology The prediction of the viral host from nucleotide sequences is split into three phases: (1) dataset curation; (2) data preprocessing; and (3) model creation and training. 2.1 Dataset curation For this research, we curated three datasets containing nucleotide sequences with host details for orthohantavirus (taxid:1980442), rabies lyssavirus (taxid:11292), and rotavirus A (taxid:28875) from the NCBI virus database [ 19 ] . While curating the dataset, the hosts with 100+ nucleotide sequences were retained. Table 1 displays an overview of these three datasets. The term “Host” refers to the host organism from which the virus was isolated, as provided by the sequence submitter [ 19 ] . Table 1: Overview of Orthohantavirus, Rotavirus A, and Rabies Lyssavirus dataset Dataset # of Hosts # of Nucleotide Sequences Orthohantavirus 9 8,575 Rabies Lyssavirus 29 22,052 Rotavirus A 12 95,197 2.2 Dataset Preprocessing Data preprocessing is a crucial part of any deep learning task to reduce future potential errors. The preprocessing task is divided into the following phases: (1) resizing the sequence length; (2) handling and encoding sequence characters. Resizing the sequence length Nucleotide sequences for viruses vary in length. To train the model, we needed to standardize the sequence length. We trained the model on lengths of 400, 600, and 1000 bases. The 1000-base length was selected as it encompassed more than half of the sequences across all datasets. However, no significant improvement was observed with longer sequences. Thus, we chose 400 bases, as it performed competitively while reducing computational costs and providing sufficient length to

Read the complete text