XAI-Driven Deep Learning for Protein Sequence Functional Group Classification

Recent research highlights a new class of attacks in federated learning that compromise model interpretability without impacting accuracy. The study reveals that adversarial clients can apply small color perturbations, shifting a model's saliency maps from meaningful regions while maintaining predictions. This method, termed the Chromatic Perturbation Module, systematically creates adversarial examples by altering color contrasts, leading to persistent poisoning of the model's internal feature attributions, challenging assumptions about model reliability.

Deep neural networks are susceptible to adversarial perturbations that can lead to incorrect predictions. The paper introduces DeepDefense, a defense framework utilizing Gradient-Feature Alignment (GFA) regularization across multiple layers to mitigate this vulnerability. By aligning input gradients with internal feature representations, DeepDefense creates a smoother loss landscape, reducing sensitivity to adversarial noise. The method shows significant robustness improvements against various attacks, particularly on the CIFAR-10 dataset.

Bladder cancer is a prevalent malignancy with a high recurrence rate of up to 78%, necessitating precise post-operative monitoring. Multi-sequence contrast-enhanced MRI is commonly utilized for recurrence detection, but interpreting these scans is challenging due to post-surgical changes. This study introduces a curated multi-sequence, multi-modal MRI dataset designed for bladder cancer recurrence prediction and proposes H-CNN-ViT, a new model aimed at enhancing prediction accuracy in this critical area.

Apo2Mol is a new diffusion-based generative framework designed for 3D molecule generation that takes into account the conformational flexibility of protein binding pockets. Traditional methods often overlook this flexibility, which can limit their effectiveness in drug discovery. Apo2Mol addresses this gap by utilizing a dataset of over 24,000 experimentally resolved apo-holo structure pairs from the Protein Data Bank, allowing for a better understanding of protein structure changes when ligands bind. This innovative approach aims to enhance the design of small molecule ligands targeting speci…

This study evaluates time-frequency feature design for binaural sound source localization (SSL), examining how feature selection affects model performance under various conditions. The research focuses on a convolutional neural network (CNN) model that utilizes combinations of amplitude-based features (magnitude spectrogram, interaural level difference - ILD) and phase-based features (phase spectrogram, interaural phase difference - IPD). Results indicate that optimal feature combinations can enhance performance more effectively than increasing model complexity.

Rice is a vital staple food globally, significantly impacting international trade, economic growth, and nutrition. Major Asian countries, including China, India, Pakistan, Thailand, Vietnam, and Indonesia, contribute to rice cultivation and utilize various rice types, such as basmati and jasmine. Ensuring the quality of rice grains is crucial for trade and reputation, but manual classification is labor-intensive and error-prone. This article explores the use of Convolutional Neural Networks (CNN) and explainable AI methods like LIME and SHAP to automate rice grain classification effectively.