Computer Science, Health Informatics, Molecular Biology, Structural Biology
9
Scopus Publications
Scopus Publications
A Study of ResNet-Enhanced CNN Versus State-of-the-Art Models for Fake News Detection P. S. S. Sumathi, V A Jisna 2025 5th International Conference on Intelligent Technologies Conit 2025, 2025 Fake news leads to dire consequences at both the individual and community levels. It has the potential to ruin reputations, incite communal riots, alter election outcomes etc. Convolutional Neural Networks (CNNs) are effective for fake news detection, but stacking many layers can limit their ability to learn high-level feature representations. Residual Networks (ResNets) allows the network to learn incremental refinements over earlier features, leading to deeper, more discriminative representations that capture the subtle patterns critical for distinguishing fake news. This study investigates the integration of ResNets into CNN models to enhance their performance in detecting fake news. Various CNN-ResNet combinations are evaluated on three publicly available datasets: MM-COVID, PubHealth, and Covid19FakeNews. The effectiveness of these models is assessed and compared with state-of-the-art methods. The ResNet-enhanced CNN model demonstrates promising results in fake news detection, though further optimization is needed for certain datasets. Future work should explore more advanced residual architectures, fine-tuning of pre-trained embeddings, and hyperparameter optimization while testing on larger, more diverse datasets to confirm real-world applicability.
Using Attention-UNet Models to Predict Protein Contact Maps V. A. JISNA, ABHAYSING PAWAR AJAY, P. B. JAYARAJ Journal of Computational Biology A Journal of Computational Molecular Cell Biology, 2024 Proteins are essential to life, and understanding their intrinsic roles requires determining their structure. The field of proteomics has opened up new opportunities by applying deep learning algorithms to large databases of solved protein structures. With the availability of large data sets and advanced machine learning methods, the prediction of protein residue interactions has greatly improved. Protein contact maps provide empirical evidence of the interacting residue pairs within a protein sequence. Template-free protein structure prediction systems rely heavily on this information. This article proposes UNet-CON, an attention-integrated UNet architecture, trained to predict residue-residue contacts in protein sequences. With the predicted contacts being more accurate than state-of-the-art methods on the PDB25 test set, the model paves the way for the development of more powerful deep learning algorithms for predicting protein residue interactions. The source codes are available in the GitHub link: (https://github.com/jisnava/UNet CON).
An End-to-End Deep Learning Pipeline for Assigning Secondary Structure in Proteins V. A. Jisna, P. B. Jayaraj Journal of Computational Biophysics and Chemistry, 2022 Protein secondary structure assignment, a subdiscipline of computational chemistry, is yet to be explored using deep learning techniques. Protein secondary structure elements are assigned to support structural analysis and prediction. Algorithms like DSSP, generally regarded as the gold standard for assigning the secondary structure of proteins, need full atom information to label protein coordinates. The PDB database has been the major repository for data on the 3D structures of proteins, nucleic acids, and other complex assemblies since 1971. However, a significant fraction of protein structures contains missing atoms. As a result, new approaches to reliably and consistently assigning secondary structures based on coarse-grained atomic coordinates are needed. While deep learning architectures have an unparalleled track record in applications such as protein structure prediction, there are only a few known deep learning solutions for structure assignment problems. While the gold standard methods are based on bonding information and other geometric characteristics, deep learning methods extract features themselves without human intervention. The benefit that standardised datasets provide to the effectiveness of deep learning systems in multiple domains motivated us to create a labeled dataset for protein structure assignment tasks. Additionally, a deep learning model, named PSSADL, solely based on [Formula: see text] coordinates was trained on the generated dataset to validate its potency. The proposed method, which combines Convolutional Neural Networks (CNN) and Long Short Term Memory (LSTM)/Bidirectional LSTM networks, has been compared to the established standards and more recent techniques. The model achieved an accuracy of [Formula: see text] on the benchmark and individual test sets. The results show that deep learning techniques have a promising future in protein structure analysis, implying that the dataset developed as part of our work will be a valuable resource for further protein structure research.
Towards building a coordinate clustered library for template-based modeling of protein structures Jisna Antony, Vishnu Sreenivas, P.B. Jayaraj 2018 IEEE Recent Advances in Intelligent Computational Systems Raics 2018, 2018 Biological activities of a human body are significantly influenced by proteins. The function of a protein can be identified by its structure. So, it is necessary to develop computational tools to predict the structure of a given protein, as experimental methods are costly and laborious. Many computational methods are available these days to predict the structure of proteins. Template-based protein structure modeling is one among such methods for protein structure prediction. Based on the assumption that similar sequences exhibit similar protein folds, template-based protein structure prediction is expected to be an accurate method to predict the structure of any protein sequence. It is a multi-step process that involves template selection, template-target sequence alignment, gap modeling and finally refining the model. Gap modeling can be effectively done now using fragment libraries. The quality of a predicted structure greatly depends on the quality of the fragment library generated. Since it has been proved that smaller length fragments model proteins more accurately than larger length fragments, the experiment was conducted with fragments of length two. In this work, a coordinate clustering approach has been introduced to reduce the conformational search space. The proposed method has been implemented and the results show a close resemblance with structures predicted by other servers. The source code of the work can be seen at: https://github.com/jisnava/Protein-Structure-Prediction
A new proposal for audio steganography in wavelet domain V.A. Jisna, C.C. Sobin Iet Conference Publications, 2012 Steganography is the art and science of writing hidden messages such that the existence of a secret communication is known only to the sender and receiver. For hiding messages different types of media are used. Audio steganography uses audio as the cover media. This paper makes a brief discussion on different audio steganography techniques. Among the techniques studied wavelet domain shows high hiding capacity and transparency. In wavelet domain different techniques are applied on the wavelet coefficients to increase the hiding capacity and perceptual transparency. The paper mainly concentrates on a survey on audio steganography in wavelet domain. Based on this survey, a proposal is made to improve the quality of retrieved data at the receiver side.
