COPAS http://localhost/xmlui/handle/123456789/1281 2026-05-17T00:17:39Z 2026-05-17T00:17:39Z Murithi, Mary Kanyiri http://localhost/xmlui/handle/123456789/6960 2026-05-12T13:16:18Z 2026-05-12T00:00:00Z

Association of CYP1B1 Gene Polymorphisms with Estrogen Receptor Positive Breast Cancer at Aga Khan University Hospital, Nairobi, Kenya Murithi, Mary Kanyiri Breast cancer is a significant global health challenge, and polymorphisms in the CYP1B1 gene have been associated with its risk. Given that the effects of genetic polymorphisms on breast cancer risk vary across populations, region-specific studies are crucial. This study assessed the associations of four CYP1B1 polymorphisms (rs10012, rs1056827, rs1056836, rs1800440) with estrogen receptor-positive breast cancer (ER+BC) risk in Kenyan women. METHODOLOGY A retrospective hospital-based case-control study involved 64 cases and 19 controls. Oversampling adjusted the case-control imbalance, increasing the control sample size to 60. DNA was extracted from buffy coat samples, and target regions were amplified and sequenced via Sanger sequencing. Sequences were analyzed using Geneious Prime for alignment, quality trimming, and SNP identification. Statistical analysis was performed using R (R 4.3.3). RESULTS The study identified significant associations between CYP1B1 polymorphisms and ER+BC risk. Specifically, the variant allele C and the codominant model (CC vs. GG) of rs10012, as well as the variant allele A, dominant (CA - CC vs. AA) and log-additive models of rs1056827, demonstrated a protective effect with ORs of 0.53 (p = 0.018, 95% CI: 0.31–0.90), 0.28 (p = 0.040, 95% CI: 0.08–0.94), 0.29 (p = 0.001, 95% CI: 0.13–0.63), 0.23 (p = 0.003, 95% CI: 0.08–0.67) and 0.51 (p = 0.005, 95% CI: 0.29–0.91), respectively. In contrast, the recessive (CC vs. GG - GC) and the log-additive models of rs10012, were linked to an increased risk of ER+BC, with ORs of 2.39 (p = 0.020, 95% CI: 1.14–5.03) and 1.97 (p = 0.014, 95% CI: 1.13–3.44), respectively. CONCLUSION These findings reveal the complex interplay between CYP1B1 polymorphisms and ER+BC risk, with some variants protecting while others increase risk. Further research is essential to fully understand the effects of these genetic variations on breasts. PhD Research Publication

2026-05-12T00:00:00Z Murithi, Mary Kanyiri http://localhost/xmlui/handle/123456789/6958 2026-05-12T12:07:16Z 2026-05-12T00:00:00Z

Polymorphisms in Estrogen Metabolizing Genes and Their Association with Estrogen Receptor-Positive Breast Cancer among patients attending Aga Khan University Hospital Nairobi, and Africa Inland Church Hospital Kijabe Murithi, Mary Kanyiri Breast cancer continues to be the most prevalent malignancy in women worldwide, with estrogen receptor-positive (ER+) tumors accounting for approximately 70% of cases. GLOBOCAN 2022 data reveal a significant global burden, with 2.3 million new cases annually, including 198,553 in Africa. Kenya reports 7,243 new cases and 3,107 deaths yearly, reflecting urgent needs for improved early detection and prevention strategies. This hospital-based case-control study aimed to determine the associations between socio-demographics, medical history, reproductive history, lifestyle factors and single nucleotide polymorphisms (SNPs) in estrogen-metabolizing genes with ER+ breast cancer risk among Kenyan women. The study compared 64 ER+ breast cancer cases with 79 benign breast disease (BBD) and 19 healthy controls from Aga Khan University Hospital Nairobi and Africa Inland Church (AIC) Kijabe Hospital. Socio-demographic and clinical data were abstracted from the questionnaires and medical records review. Estrogen plays a pivotal role in the pathogenesis of ER+ breast cancer. Individual genetic variation in estrogen-metabolizing enzymes can significantly alter the production, activity, and clearance of estrogen and its metabolites, thereby modifying cancer risk and progression. Single nucleotide polymorphisms (SNPs) in genes such as CYP1A1, CYP1B1, CYP3A5, and COMT can lead to differential enzymatic activity, influencing the critical balance between carcinogenic catechol estrogens and their detoxified forms. Consequently, these SNPs are considered key biomarkers for understanding interindividual susceptibility and prognosis in ER+ breast cancer. Five functionally relevant single nucleotide polymorphisms (SNPs) in key estrogen metabolism genes including; rs4646903 and rs1048943 of CYP1A1; rs1056836 of CYP1B1; rs776746 of CYP3A5 and rs4680 of COMT were analyzed using polymerase chain reaction restriction fragment length polymorphism (PCR-RFLP). Four additional SNPs (rs10012, rs1056827, rs1056836 and rs1800440 of CYP1B1) were analyzed via Sanger sequencing methods. Key findings demonstrated that women aged 50 years and older and postmenopausal women faced significantly elevated breast cancer risk. Genetic analysis revealed complex patterns: the alternative C allele of rs4646903 in CYP1A1 showed a protective effect against ER+ cancer development (OR=0.44, 95% CI [0.19-0.99], p = 0.048), but was also paradoxically associated with increased risk of aggressive Luminal B subtypes (OR=3.83, 95% CI [1.35-10.84]). Two CYP1B1 variants – alternative C allele in rs1056836 (OR = 0.34, 95% CI [0.19–0.62], p = 0.0003) and alternative A allele in rs1056827 (OR=0.43, 95% CI [0.19-0.98], p = 0.045) – were associated with reduced likelihood of malignant transformation from benign breast disease. Linkage disequilibrium analysis revealed strong association between rs10012 and rs1056827 (D' = 0.9466, r² = 0.5767, p = 1.08 × 10⁻¹³), confirming that these loci form a haplotype block. Haplotype frequency analysis identified eight distinct CYP1B1 haplotypes, with the G–A–C–T haplotype (R–S–V–N) predominating in benign samples (42.3%) suggesting a protective role, while the G–C–C–T haplotype (G–A–V–N) showed elevated frequency in cases (12.5%) compared to controls (8.3%), indicating potential risk association. Tertiary structure modeling using AlphaFold Server revealed that the four polymorphic residues—R48G, A119S, L432V, and N453S—occupy functionally distinct domains: R48 in the N-terminal membrane interaction region, A119 in the structural core, L432 near the substrate-binding pocket, and N453 in the heme-binding region. Comparative structural analysis of the wild-type (R–A–L–N), reference (G–S–V–N), risk-associated (G–A–V–N), and protective (R–S–V–N) haplotypes demonstrated that the protective haplotype uniquely retains the wild-type R48 residue while incorporating S119 and V432, suggesting that proper membrane anchoring may be critical for maintaining protective function. Conversely, the risk-associated haplotype combines loss of R48 with the high-activity V432 variant, potentially synergistically enhancing carcinogenic estrogen metabolism. Ramachandran plot analysis confirmed structural reliability, with >90% of residues in favored regions across all models. These findings provide important insights into the complex interplay between socio-demographic factors, genetic predisposition, and structural consequences of CYP1B1 haplotypes in ER+ breast cancer development among Kenyan women. The study's findings underscore the importance of developing population-specific risk assessment tools that incorporate both genetic and structural information to combat Kenya's disproportionate breast cancer mortality burden. Future research should focus on molecular dynamics simulations to elucidate the dynamic mechanisms underlying haplotype-specific functional differences, validation in larger cohorts, and exploration of potential clinical applications in preventive strategies and personalized treatment approaches. PhD in Biochemistry

2026-05-12T00:00:00Z Hakizimana, Leopord http://localhost/xmlui/handle/123456789/6955 2026-05-12T08:23:09Z 2026-05-11T00:00:00Z

A Hybrid-Based Classification and Regression Model for Predicting Malaria Outbreak Hakizimana, Leopord Malaria outbreaks remain a main public health challenge worldwide, particularly in Sub Saharan Africa. Fast and correct forecast of malaria outbreaks is critical for permitting timely interventions, decreasing morbidity and death, and ensuring effective sharing of limited healthcare resources. In the last ten years, Data mining and machine learning methods gained widespread attention in complex prediction tasks, such as healthcare analytics, financial and environmental monitoring prediction. Regardless of these improvements, existing malaria outbreak forecast models frequently show limitations in accuracy, adaptability, and applied usefulness. Numerous existing approaches depend solely on either regression or classification approaches, which limits their ability to attain the complex and dynamic interactions between environmental, climatic, and epidemiological factors that affect malaria transmission. This research introduces a new hybrid based predictive model that mix both regression and classification methods in a two-phase framework, marking to enhance the accuracy and reliability of malaria outbreak predictions. The first phase applies a regression model to predict the expected number of malaria cases by examining historical epidemiological data, climate variables and other appropriate environmental indicators. The second phase applies a classification model to determine the likelihood of an outbreak occurring within a given region and time frame, transforming quantitative predictions into actionable early warning signals. Through amalgamation these supplementary methods, the hybrid model influences the strengths of both regression and classification, outcome of in enhanced prediction performance, robustness, and adaptability under diverse outbreak situations. Comprehensive experimentations were done using publicly accessible and region-specific malaria datasets, and the outcomes show that the hybrid model significantly outperforms conventional single-method approaches. The attained model predictive accuracies of 96% through training and 93% in testing, demonstrating strong generalizing capabilities. Likewise, the hybrid approach improves the decision-making aptitudes of healthcare systems by providing timely and reliable information that support evidence-based interventions, such as targeted mosquito control, resource prioritization and prophylactic measures. Study has important inferences for health professionals and authorities, policymakers, and international health organizations endeavoring to reduce malaria burden efficiently. The research helps to healthcare and machine learning fields by giving a scalable and adaptable framework for disease outbreak prediction. It also serves as a foundation for future studies on hybrid and ensemble Machine Learning models, mostly in the perspective of infectious diseases prediction. Future work is endorsed to explore the combination of supervised and unsupervised hybrid methods, integration of real-time epidemiological and climatic data streams, and assessment under large-scale, dynamic outbreak situations. Further, this study emphasizes the capability of hybrid machine learning models to renovate disease outbreak prediction by merging the predictive strengths of regression and classification methods. The findings show the importance of adapting data driven strategies for enabling early detection, timely interventions, public health preparedness and ultimately, the reduction of malaria transmission and its associated health and socioeconomic effects xviii Keywords: Hybrid-Based Model, Classification, Regression, Malaria, Outbreak Prediction PhD in Computer Science

2026-05-11T00:00:00Z Ngao, Narshion Matai http://localhost/xmlui/handle/123456789/6945 2026-05-11T08:30:09Z 2026-05-11T00:00:00Z

A Deep Neural Network Model for Detection of Urgency in the Short Message Services Ngao, Narshion Matai Timely identification of urgent patient messages is critical for effective clinical decision-making in mobile health (mHealth) programs, particularly in low-resource settings where healthcare workers manage large volumes of incoming short message service (SMS) communication. In Kenyan maternal and child health programs, nurses manually triage multilingual patient messages, a process that contributes to delayed responses and increased risk of missed urgent cases. This study investigates the effectiveness of contextual natural language processing (NLP) models for automatically classifying patient SMS messages into urgency categories within a real-world mHealth environment. Using a dataset of 11,129 manually labelled multilingual SMS messages from 772 participants enrolled in the Mobile Solutions for Women and Children’s Health (Mobile WACh NEO) program in Kenya, urgency detection was formulated as a supervised binary classification task aligned with clinical triage workflows. Baseline models employing unigram and bigram features with penalized logistic regression were compared against contextual embedding approaches, including multilingual BERT (mBERT), SwahBERT, and AfriBERT. Transformer models were adapted to the clinical domain through domain-specific pretraining and task-adaptive fine-tuning. To mitigate contextual sparsity inherent in short SMS messages, prior nurse or system messages were concatenated with the current message to form context-aware input representations. Model development followed explicit train, development, and test splits, with cross-validation applied during training to support robust model selection and reduce overfitting. Performance was evaluated using precision, recall, and F1-score, emphasizing clinical utility for both triage and prioritization objectives. Transformer architectures substantially outperformed frequency-based baselines, achieving F1 improvements of up to 0.186 relative to bigram models. Our best performing model was mBERT model pretrained on task-level adaptation using nurse context before fine-tuning. This model got a precision of 50%, recall of 45% and F1 score of 47%, which were below the thresholds we set for either a triage model or prioritization model. However, incorporating nurse conversational context reduced performance gaps between configurations (e.g., ΔF1 decreasing from approximately 0.080 in non-contextual mBERT to 0.032 with nurse context), while task-adaptive pretraining provided incremental yet consistent gains. Although performance did not fully meet predefined clinical usefulness thresholds, context-aware fine-tuned transformer models demonstrated improved recall, indicating reduced risk of missed urgent messages. Overall, the findings confirm that contextual transformer-based models offer meaningful advantages over traditional representations in multilingual, low resource clinical SMS environments. While additional advances in architecture and domain adaptation are needed to reach optimal deployment standards, the results align with contemporary state-of-the-art NLP practices and support the feasibility of automated decision-support tools to augment nurse triage workflows in mHealth systems. Keywords: Urgency Detection, Contextual NLP, Multilingual Transformers, mHealth, Clinical Decision Support MSc in Computer Systems

2026-05-11T00:00:00Z