Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Candra Zonyfar has a total of 44 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    DeepCPP: A Deep Learning‐Based Framework for Enhanced Prediction of Cell Penetrating Peptides

    by Soualihou Ngnamsie Njimbouom, Candra Zonyfar, Jeong‐Dong Kim

    Cell‐penetrating peptides (CPPs) are bioactive molecules characterized by a positive charge, primarily attributed to lysine, arginine, or histidine residues, which enable their traversal across cells’ lipid bilayer membranes. This unique property positions CPPs as promising candidates for intracellular delivery of therapeutic agents. The expanding database of experimentally validated CPPs has facilitated significant advancements in computational modeling, enabling in silico prediction and engineering of established and novel CPPs. Despite progress in this domain, challenges persist in improving the performance of CPP prediction models. This study introduces DeepCPP, a new deep learning‐based framework for efficient CPP prediction. DeepCPP employs a multihead decision‐level late‐fusion approach that uses a trainable weighted fusion mechanism to combine branch‐specific predictions generated by a multilayer convolutional neural network‐guided encoder and a multilayer graph neural network‐guided encoder. These encoders learn evolutionary features from an evolutionary information‐based representation and chemical, topological, and stereochemical features from a molecular graph‐based representation, respectively. Comprehensive experiments conducted on established benchmark datasets reveal the effectiveness of the DeepCPP model, exhibiting performance superior to state‐of‐the‐art models in CPP prediction. To promote broader accessibility, we provide a user‐friendly web service for the proposed DeepCPP model, accessible at https://deepcpp.streamlit.app/ .

  • Open Access

    Transfer learning-based malnutrition classification using VGG16 and comparative analysis of CNN architectures

    by Ahmad Fauzi, Haerul Yuda Aditiya, Maharina Maharina, Sihabudin Sihabidin, Muhammad Ansari Adista, Iflan Naufal, Natasya Eka Nanda Sonia Puri, Candra Zonyfar

    Childhood malnutrition remains a critical public health challenge in developing countries, with Indonesia ranking fifth globally for stunting burden. Traditional anthropometric assessment methods are time-consuming, resource-intensive, and require trained personnel, necessitating efficient computer-based early detection approaches. This study proposes a deep learning-based method for automated nutritional status classification using facial image analysis. We developed and compared multiple transfer learning architectures including visual geometry group 16 (VGG16), densely connected convolutional network 121 (DenseNet121), mobile network version 2 (MobileNetV2), and residual network 50 (ResNet50) for classifying children into three categories: healthy, malnutrition, and stunting. Results demonstrated that VGG16, a simpler architecture trained for only 10 epochs, achieved optimal performance with 91.8% accuracy, and significantly outperforming more complex modern architectures like ResNet50. This finding challenges the conventional assumption that newer, deeper models invariably perform better, and particularly when working with limited medical datasets. The study revealed that longer training durations led to performance degradation due to overfitting, emphasizing the importance of balancing model complexity with dataset characteristics. These findings support the development of practical artificial intelligence (AI)-based malnutrition screening systems suitable for resource-constrained environments, potentially improving early detection capabilities, and public health outcomes in developing regions.

  • Open Access

    Profiling numeracy literacy among ninth-grade students: Empirical evidence from junior secondary education

    by Andes Safarandes Asmara, Yusfita Yusuf, Anggy Giri Prawiyogi, Candra Zonyfar, Muhammad Akmal Alhamssyah, Muhammad Naufal Alfarid

    Numeracy literacy in this study refers to students' capacity to interpret quantitative information in context, mathematize situations, use representations, justify their reasoning, select strategies, and employ symbols and tools to support defensible decisions. To address Indonesia's low performance on PISA tasks, we developed an operational profile of 9th-grade students using OECD indicators adapted to the Indonesian junior high school geometry curriculum through Candi Jiwa-based items. Using a descriptive qualitative design with quantitative support, we administered seven open-ended geometry problems to students in Karawang, Purwakarta, and Subang. Test scores were summarized to classify achievement levels. At the same time, classroom observations and semi-structured interviews with students and teachers were used to diagnose reasoning and strategy tendencies and recurring error patterns. Results show that 85.71% of students scored below 60 (low), 14.28% scored 60–80 (medium), and none exceeded 80 (high). Students' strongest area was written communication, mainly descriptive or procedural (restating givens and listing steps). Weaknesses clustered in mathematization, reasoning and argumentation, representation, strategy selection, and use of formal symbols and mathematical tools, indicating procedural competence with limited transfer to contextual modeling. Candi Jiwa serves as a place-based anchor that can lower entry barriers and guide culturally grounded, cognitively manageable task design aligned with OECD processes.


  • Open Access

    R2eGIN: Residual Reconstruction Enhanced Graph Isomorphism Network for Accurate Prediction of Poly (ADP–Ribose) Polymerase Inhibitors

    by Candra Zonyfar, Soualihou Ngnamsie Njimbouom, Sophia Mosalla, Jeong-Dong Kim

    An advanced graph neural network (GNN) is of great promise to facilitate predicting Poly ADPribose polymerase inhibitors (PARPi). Recent studies design models by leveraging graph representations and molecular descriptor representations, unfortunately, still face challenges in comprehensively capturing spatial relationships and contextual information between atoms. Moreover, combining molecular descriptors with graph representations may introduce information redundancy or lead to the loss of intrinsic molecular structures. To this end, we proposed a novel Residual Reconstruction Enhanced Graph Isomorphism Network (R2eGIN) learning model. Specifically, we first designed a residual GIN to learn molecular representations, reduced the impact of vanishing gradients, and enabled the model to capture long-range dependencies. Then, the reconstruction block, by predicting adjacency matrices and node features, was adopted to reconstruct the input graph. To prove the effectiveness of the proposed model, extensive experiments were conducted on 4 data sets of PARPi and compared with 7 existing models. Our evaluation of R2eGIN, conducted using 4 PARPi data sets, shows that the proposed model is comparable to or even outperforms other state-of-the-art models for PARPi prediction. Furthermore, R2eGIN can revolutionize the drug repurposing process through a substantial reduction in the time and costs commonly encountered in traditional drug development methods.


  • Open Access

    GTransCYPs: an improved graph transformer neural network with attention pooling for reliably predicting CYP450 inhibitors

    by Candra Zonyfar, Soualihou Ngnamsie Njimbouom, Sophia Mosalla, Jeong-Dong Kim

    State‑of‑the‑art medical studies proved that predicting CYP450 enzyme inhibitors is beneficial in the early stage of drug discovery. However, accurate machine learning-based (ML) in silico methods for predicting CYP450 inhibitors remains challenging. Here, we introduce GTransCYPs, an improved graph neural network (GNN) with a transformer mechanism for predicting CYP450 inhibitors. This model significantly enhances the discrimination between inhibitors and non-inhibitors for five major CYP450 isozymes: 1A2, 2C9, 2C19, 2D6, and 3A4. GTransCYPs learns information patterns from molecular graphs by aggregating node and edge representations using a transformer. The GTransCYPs model utilizes transformer convolution layers to process features, followed by a global attention-pooling technique to synthesize the graph-level information. This information is then fed through successive linear layers for final output generation. Experimental results demonstrate that the GTransCYPs model achieved high performance, outperforming other state-of-the-art methods in CYP450 prediction. Scientific contribution The prediction of CYP450 inhibition via computational techniques utilizing biological information has emerged as a cost-effective and highly efficient approach. Here, we presented a deep learning (DL) architecture based on GNN with transformer mechanism and attention pooling (GTransCYPs) to predict CYP450 inhibitors. Four GTransCYPs of different pooling technique were tested on an experimental tasks on the CYP450 prediction problem for the first time. Graph transformer with attention pooling algorithm achieved the best performances. Comparative and ablation experiments provide evidence of the efficacy of our proposed method in predicting CYP450 inhibitors. The source code is publicly available at https://github.com/zonwoo/GTransCYPs .


  • Open Access

    ERL‐ProLiGraph: Enhanced representation learning on protein‐ligand graph structured data for binding affinity prediction

    by Gloria Geine Paendong, Soualihou Ngnamsie Njimbouom, Candra Zonyfar, Jeong‐Dong Kim

    Predicting Protein‐Ligand Binding Affinity (PLBA) is pivotal in drug development, as accurate estimations of PLBA expedite the identification of promising drug candidates for specific targets, thereby accelerating the drug discovery process. Despite substantial advancements in PLBA prediction, developing an efficient and more accurate method remains non‐trivial. Unlike previous computer‐aid PLBA studies which primarily using ligand SMILES and protein sequences represented as strings, this research introduces a Deep Learning‐based method, the Enhanced Representation Learning on Protein‐Ligand Graph Structured data for Binding Affinity Prediction (ERL‐ProLiGraph). The unique aspect of this method is the use of graph representations for both proteins and ligands, intending to learn structural information continued from both to enhance the accuracy of PLBA predictions. In these graphs, nodes represent atomic structures, while edges depict chemical bonds and spatial relationship. The proposed model, leveraging deep‐learning algorithms, effectively learns to correlate these graphical representations with binding affinities. This graph‐based representations approach enhances the model‘s ability to capture the complex molecular interactions critical in PLBA. This work represents a promising advancement in computational techniques for protein‐ligand binding prediction, offering a potential path toward more efficient and accurate predictions in drug development. Comparative analysis indicates that the proposed ERL‐ProLiGraph outperforms previous models, showcasing notable efficacy and providing a more suitable approach for accurate PLBA predictions.


  • Open Access

    CAB‐Net: Convolutional attention BLSTM network for enhanced leukocyte classification

    by Candra Zonyfar, Jeong‐Dong Kim

    In recent years, deep learning techniques have been increasingly utilized to automate and reduce human errors in leukocyte classification. However, reliability and accuracy issues remain a challenge. This paper proposes CAB‐Net, an end‐to‐end deep learning‐based model that integrates a convolutional neural network with an attention mechanism and BLSTM. The model uses a stack of convolutional and pooling layers to extract initial features from input images, which are then fed into the attention block to generate class‐specific discriminative features and BLSTM employed to model deterministic class correlations between highlighted discriminative features and labels. Experimental results show that CAB‐Net outperforms state‐of‐the‐art methods in term of accuracy, precision, recall, F1‐Score, and AUC‐ROC. In addition, the CAB‐Net model demonstrates good performance even with low‐resolution images. Consequently, CAB‐Net can be considered as the current state‐of‐the‐art prediction model for improving leukocyte classification in clinical environments.


  • An ensemble learning-based machine learning with voting mechanism for chronic disease prediction

    by Soualihou Ngnamsie Njimbouom, Jeong-Dong Kim, Candra Zonyfar, Hyun Lee, Taek Lee, Jung-Been Lee

    Chronic diseases present a significant challenge in healthcare, often requiring ongoing medical attention and posing limitations on patients' daily activities. Diagnosis of such diseases is hindered by the absence of specific symptoms making them hard to detect and or prevent. Addressing this issue, researchers have turned to computational approaches, analyzing patients' medical records to predict the presence or absence of chronic diseases with promising results. However, there is potential for further improvement. This paper introduces an ensemble classifier, ELVot-CroDiP, designed to enhance chronic disease prediction. ELVot-CroDiP harnesses the collective strengths of various machine learning algorithms through a majority voting mechanism, thereby boosting predictive accuracy. The model’s performance was evaluated using precision, recall, f1-score, and accuracy across five medical datasets, including heart disease, chronic kidney disease, diabetes, and heart stroke. Comparative analysis shows that ELVot-CroDiP offers superior performance in predicting chronic diseases compared to existing models, as evidenced by its higher scores in the mentioned evaluation metrics for each dataset tested.


  • Open Access

    HCNN-LSTM: Hybrid Convolutional Neural Network with Long Short-Term Memory Integrated for Legitimate Web Prediction

    by Candra Zonyfar, Jung-Been Lee, Jeong-Dong Kim

    Phishing techniques are the most frequently used threat by attackers to deceive Internet users and obtain sensitive victim information, such as login credentials and credit card numbers. So, it is important for users to know the legitimate website to avoid the traps of fake websites. However, it is difficult for lay users to distinguish legitimate websites, considering that phishing techniques are always developing from time to time. Therefore, a legitimate website detection system is an easy way for users to avoid phishing websites. To address this problem, we present a hybrid deep learning model by combining a convolution neural network and long short-term memory (HCNN-LSTM). A one-dimensional CNN with a LSTM network shared estimation of all sublayers, then implements the proposed model in the benchmark dataset for phishing prediction, which consists of 11430 URLs with 87 attributes extracted of which 56 parameters are selected from URL structure and syntax. The HCNN-LSTM model was successful in binary classification with accuracy, precision, recall, and F1-score of 95.19%, 95.00%, 95.00%, 95.00%, successively outperforming the CNN and LSTM. Thus, the results show that our proposed model is a competitive new model for the legitimate web prediction tasks.


  • RD-Classifier: Reduced Dimensionality Classifier for Alzheimer’s Diagnosis Support System

    by Soualihou Ngnamsie Njimbouom, Gelany Aly Abdelkader, Candra Zonyfar, Hyun Lee, Jeong-Dong Kim