4th International Conference on Electrical Engineering (ELEE 2024)

December 28 ~ 29, 2024, Dubai, UAE

Accepted Papers


Multi-domain Interactions of Tph2 Predicted by Alphafold 3

Yasemin Ocak1, & Jason O. Matos2, 1Ridge High School, Basking Ridge, New Jersey, USA, 2Institute for Plant-Human Interface, Northeastern University, Boston, MA 02120, USA

ABSTRACT

Serotonin is an important neurotransmitter in learning, memory, happiness, as well as regulating homeostasis. Abnormal levels of serotonin can result in disorders such as depression, OCD, PTSD, schizophrenia, and serotonin syndrome. Tryptophan hydroxylase 2 (TPH2) catalyzes the rate-limiting step in the production of serotonin. Mutations in TPH2 can result in the disorders discussed above. Thus, TPH2 is a major drug target. TPH2 functions as a homotetramer in the cell, and consists of catalytic, regulatory, and oligomeric domains. Details of how these domains interact are lacking. Structural insight into TPH2 could help find molecules that regulate its activity. While we do know the structures of the domains separately, we don’t know exactly how they interact. In order to determine how these domains interact, the new technology AlphaFold 3 was used. While AlphaFold 3 is said to have transformed the field of structural biology, in this study we will test whether AlphaFold predictions are accurate by proposing mutants of TPH2 that are at the interface of the regulatory and catalytic domains. Taken together, our ef orts will be able to determine if the structure predicted by AlphaFold for a multi- domain protein is accurate.

Keywords

Serotonin, Tryptophan hydroxylase 2 (TPH2), AlphaFold 3, Depression, OCD, PTSD, Serotonin syndrome, Homotetramer, Catalytic domain, Regulatory domain, Domain interactions, Structural biology, Drug target, Protein structure prediction, TPH2 mutations.


Diffusion and Gan-based Cross-modal Synthesis for Enhanced 3d Medical Imaging: From 2d X-ray to High-fidelity 3d Volumetric Reconstructions

Elsun Nabatov, School of Computing, Newcastle University, Newcastle Upon Tyne, United Kingdom

ABSTRACT

This study introduces a novel hybrid framework combining Generative Adversarial Networks (GANs) and diffusion models to transform 2D X-ray images into high-fidelity 3D volumetric reconstructions. Addressing challenges such as high costs, radiation exposure, and limited access to 3D imaging modalities like MRI and CT, the framework balances computational efficiency and output fidelity. GANs enable rapid synthesis, while diffusion models ensure refined detail, validated on benchmark datasets (NIH Chest X-ray, LIDC-IDRI, BraTS). The framework achieved significant improvements in structural and perceptual quality, with SSIM of 0.92 and PSNR of 35.2 dB. Key applications include enhanced diagnostic imaging in resource-constrained settings and reduced reliance on highdose modalities. This approach democratizes access to advanced imaging technologies, offering a cost-effective, lowradiation alternative for global biomedical research and clinical diagnostics.

Keywords

Cross-modal synthesis, Diffusion models, GANs, 3D medical imaging, 2D-to-3D transformation.


Artificial Intelligence in Determining Optimal Questions in Assessing Social Economic Status of Individuals for Routine Immunization Services in Tanzania

Deogratias Mzurikwao1,3, Lwidiko Edward Mhamilawa1, Daudi Simba1, Belinda Balandya1, Evelyne Assenga1, Charles Okanda Nyatega2, Jonathan Zeramula4, Seif Wibonela1, Zacharia Mzurikwao2, Bruno Sunguya1, 1Muhimbili University of Health and Allied Sciences, 2Mbeya University of Science and Technology, 3Emerging Technologies for Health lab (ETH)-MUHAS, 4Tanzania Atomic Energy Commission (TAE)

ABSTRACT

This study aimed to determine the lowest optimal questions that could accurately determine the socio-economic status (SES) score of the participants and determine their validity when compared to the standard wealth index. Principal Component Analysis (PCA), Convolutional Neural Networks (CNN) and Artificial Neural Networks (ANN) techniques were applied using DHS wealth index as the gold standard. Eight DHS questions were found to be optimal for assessing household’s SES with high sensitivity (76.9%) and specificity (94.2%). The correlation with DHS standard wealth index was R2 = 0.76. The study has also shown the potential for using CNN as a method to identify valid questions that can be applied in other domains. Our findings open the possibility of using SES as one of the factors to determine access and completion of routine immunization services. This is important in identifying and targeting populations at risk to enable focussed interventions to increase vaccine coverage.

Keywords

Artificial Intelligence, Social economic status, PCA, ANN, CNN.