http://localhost/xmlui/handle/123456789/1281 COPAS 2026-04-05T15:22:45Z http://localhost/xmlui/handle/123456789/6930 Mutational and Phylogenetic Analyses of Antifolate and Artemisinin Resistance in Plasmodium falciparum Dried Blood Spots Obtained from Patients Attending Three Hospitals of Eritrea Mukhongo, Harriet Natabona Malaria causes approximately 200 million infections and 500,000 deaths yearly. The emergence and spread of antimalarial drug resistance is a major challenge towards control efforts globally. Nonetheless, integrated vector control methods have reduced malaria transmission, resulting in decreased human-parasite reservoirs and lessened spread of antimalarial drug resistance. Eritrea, located in the horn of Africa, has witnessed a considerable reduction in malaria deaths from 405 to 4 between 1998 and 2023. This is attributed to a combination of community health campaigns, prompt case management, and integrated vector management. In Eritrea, Sulfadoxine-Pyrimethamine was previously used as a first-line antifolate treatment, and currently, artesunate is the first-line artemisinin treatment. Due to limited molecular data on these antifolate and artemisinin treatments, the first objective of this study, was to determine antifolate resistance-associated genetic mutations in codon position K540E of Pf-DHPS gene, and codon positions N51I, C59R and S108N of Pf-DHFR gene. The second objective, was to determine artemisinin resistance-associated genetic mutations in codon positions Y493H, R539T, I543T, and C580Y of Pf-K13 gene. The third objective, was to determine the phylogenetic relationships between the genetic markers sampled in this study, and corresponding globally identified genetic markers from other studies. Sample size was determined using Fisher’s formula, and based on patient availability, inclusion and exclusion criteria. Nineteen dried blood spot samples were collected from patients infected with P. falciparum mono-infection, visiting Adi Quala, Keren, and Gash Barka hospitals. Genomic DNA extraction, nested-PCR amplification and Sanger-sequencing of Pf-DHFR, Pf-DHPS, and Pf-K13 partial gene regions was achieved for nine dried blood spots. Sequence contig assembly, genetic mutation visualization, and phylogenetic analyses were performed in CLC main workbench v21.0.4, Jalview v2.11.1.4, and MEGAv7.0. Mutational analyses identified the single-mutant K540E of Pf-DHPS in Adi Quala (n=1), Keren (n=1), and Gash Barka (n=1); double-mutant N51I+S108NI of Pf-DHFR in Adi Quala (n=2); triple-mutant S108N+C59R+N51I of Pf-DHFR in Keren (n=1); mixed-mutant (S108N+N51I+K540E) of Pf-DHFR and Pf-DHPS in Gash Barka (n=1). These findings suggested the likely presence of the quintuple-mutant (S108N, C59R, N51I+A437G, K540E) of Pf-DHFR and Pf-DHPS, associated with full resistance, and used to predict Sulfadoxine-Pyrimethamine treatment failure. Mutational analyses of Pf-K13 identified wild-type haplotypes of Y493Y+R539R+I543I+C580C in Adi Quala (n=2) and wild-types of C580C in Keren (n=1) and Gash Barka (n=3). These findings suggested the likely absence of artemisinin resistance, and predicted artesunate was still effective for malaria treatment. The Dhfr phylogeny predominantly identified the double-mutant haplotype (N51I + S108N) at an estimated distribution ranging from low to high prevalence in Western Kenya (p=0.3%), Myanmar (p=2.5%), India (p=7%) and Sudan (p=80%). The K540E mutation was predominantly identified in the Dhps phylogeny, at an estimated distribution ranging from low to moderate prevalence in Ghana (p=3.4%), Equatorial Guinea (p=5.1%), and Sudan (p=65.7%). These analyses suggested a low to moderate spread of antifolate resistance. The K13 phylogeny predominantly identified wild-type haplotype (Y493Y+R539R +I543I +C580C) at an estimated distribution of high prevalence in Ghana (p=100%), Nigeria (p=96.9%), Niger (p=90%) and Angola (p=>80%). This suggested a limited spread of artemisinin resistance. Future recommendations from this study should estimate the mutational prevalence and phylogenetic distribution of antifolate and artemisinin resistance in a larger sample population. MSc in Bioinformatics and Molecular Biology 2026-03-31T00:00:00Z http://localhost/xmlui/handle/123456789/6926 An Enhanced K-Means Clustering Information Mining Model for Selective Dissemination of Information for Library Users Too, Titus Kiprugut Rotich The amount of information materials held by academic libraries is enormous and ever increasing at an astonishing rate as new pieces of information are now not only present in the physical form but also in digital sources. These large numbers of information resources are a challenge to librarians’ on how they can effectively and efficiently provide such relevant information resources to users. This abundance of information has created a twofold challenge. First, users have to navigate through huge information to locate what they need and that is relevant. Secondly, the challenge in computational problems where most human errors go unidentified and corrected. Mining of information is essential, as looking for data is in itself a troublesome process. Clustering is concerned with the grouping of unlabeled feature vectors into clusters, such that samples within a cluster are more similar to each other than samples belonging to different clusters. Usually, it is assumed that the number of clusters is known in advance, but otherwise no prior information is given about the data. Clustering can be used for information mining in the library. K-means is an algorithm for clustering a set of unlabeled feature vectors X: {x1, …, xn} that are drawn independently from the mixture density p(X|θ) with a parameter set θ. The main objective of this study implemented and evaluated a clustering mining model for selective dissemination of information at an academic library. The research methodology approach that was used was quantitative experimental research design. The dataset population was obtained from an online open-source repository containing datasets that was acquired by scrapping goodreads.com. That dataset was 48 MB in size. The research was based on purposive sampling technique, a form of non-probability sampling. The K-means model for clustering used a set of unlabeled feature vectors X:{x1, … , xn} that were drawn independently from the mixture density p(X|θ) with a parameter set θ. The dataset was divided into two-dimensional with P = 12 data points naturally clustered into K = 3 clusters. The dataset was imported online from a CSV file using python import library function numpy. The dataset contained both training and test data that provided an enhanced validation of the model. Once the training dataset was created, it was time to train the model. AutoML Google Colab was used to train the model that got an RMSE of 0.198. AutoML used a number of sophisticated models such as neural architecture search, which build a learning networks one layer at a time. The comparison results from the experiments prove that the implemented recommendation k-mean clustering model approach for clustering information mining model to enhance selective dissemination of information for library users has the least percentage of mean vector and covariance matrix which resulted in a higher accuracy of 71%. When matrix vector and covariance vector are low, it brings an impression of an efficient model approach to clustering information mining model. In future, work may be extended by adding suitable pre-processing approaches to improve the datasets as well as features selection approach to improve the classification accuracy. Future work should also extend on time series dynamic data that are in real time, thereby developing new technique against improved hybrid approaches. MSc in Information Technology 2026-03-26T00:00:00Z http://localhost/xmlui/handle/123456789/6920 A Modified Hidden Markov Model for Predicting Cardholder Purchasing Patterns Across Multichannel Transactions with Explainability Okoth, Jeremiah Otieno One of the biggest challenges in the card payments industry is figuring out the purchasing intent of cardholders and key transaction drivers. Prior studies have mostly focused on e-commerce, credit card transactions, and fraud detection, often overlooking other card kinds and different merchant categories. To fill the identified gap, this study utilized a sequential card transactional dataset to analyze cardholder purchasing patterns within four merchant acquiring sectors: restaurants, health care facilities, fuel stations, and social joints. The main objective of this study was to construct a predictive model that can accurately profile cardholders and detect their intent beyond the horizon of fraud detection. Unlike conventional methods such as Naïve Bayes, decision trees, and support vector machines, the proposed model dynamically represents transactional behavior using the Hidden Markov Model. To achieve resilience and flexibility, the methodology employed three HMM problems, i.e., initialization, decoding, and evaluation. Additionally, performance optimization techniques such as feature engineering, principal component analysis (PCA), sensitivity analysis, and 5-fold cross-validation were employed. By integrating the capabilities of the surrogate decision tree model with principal component analysis-transformed Hidden Markov Model outputs, this research introduced a novel computational breakthrough that generates an interpretable framework, linking the predictive power of the Hidden Markov Models with stakeholders' decision-making requirements. This hybrid approach potentially overcomes a major drawback of opaque sequential modeling techniques by enabling stakeholders to understand both the anticipated purchasing behaviors and the causes of particular transactional patterns that lead to specific consumer intent classifications. With 100% accuracy and precision, 99% recall, a 98.5% F1-score, and a ROC-AUC of 0.992, the experimental results exhibited outstanding performance. Conventional models like SVM, decision trees, Naïve Bayes, transformer-based models, and LSTM networks are outperformed by the results. Despite the encouraging outcomes, the study acknowledged a number of important limitations. Four merchant categories are the only ones included in the dataset, which may limit its applicability to the larger payments ecosystem. The near-perfect performance metrics were largely driven by the extensive optimization pipeline, and could potentially pose significant risks of overfitting. Future studies ought to use more broadly applicable methodologies and cover a greater variety of merchant sectors. This study demonstrates how well HMMs predict cardholder behavior, offering merchants and stakeholders insightful information. MSc in Computer Systems 2026-03-18T00:00:00Z http://localhost/xmlui/handle/123456789/6918 A Modified Hidden Markov Model for Predicting Cardholder Purchasing Patterns Across Multichannel Transactions with Explainability Okoth, Jeremiah Otieno One of the biggest challenges in the card payments industry is figuring out the purchasing intent of cardholders and key transaction drivers. Prior studies have mostly focused on e-commerce, credit card transactions, and fraud detection, often overlooking other card kinds and different merchant categories. To fill the identified gap, this study utilized a sequential card transactional dataset to analyze cardholder purchasing patterns within four merchant acquiring sectors: restaurants, health care facilities, fuel stations, and social joints. The main objective of this study was to construct a predictive model that can accurately profile cardholders and detect their intent beyond the horizon of fraud detection. Unlike conventional methods such as Naïve Bayes, decision trees, and support vector machines, the proposed model dynamically represents transactional behavior using the Hidden Markov Model. To achieve resilience and flexibility, the methodology employed three HMM problems, i.e., initialization, decoding, and evaluation. Additionally, performance optimization techniques such as feature engineering, principal component analysis (PCA), sensitivity analysis, and 5-fold cross-validation were employed. By integrating the capabilities of the surrogate decision tree model with principal component analysis-transformed Hidden Markov Model outputs, this research introduced a novel computational breakthrough that generates an interpretable framework, linking the predictive power of the Hidden Markov Models with stakeholders' decision-making requirements. This hybrid approach potentially overcomes a major drawback of opaque sequential modeling techniques by enabling stakeholders to understand both the anticipated purchasing behaviors and the causes of particular transactional patterns that lead to specific consumer intent classifications. With 100% accuracy and precision, 99% recall, a 98.5% F1-score, and a ROC-AUC of 0.992, the experimental results exhibited outstanding performance. Conventional models like SVM, decision trees, Naïve Bayes, transformer-based models, and LSTM networks are outperformed by the results. Despite the encouraging outcomes, the study acknowledged a number of important limitations. Four merchant categories are the only ones included in the dataset, which may limit its applicability to the larger payments ecosystem. The near-perfect performance metrics were largely driven by the extensive optimization pipeline, and could potentially pose significant risks of overfitting. Future studies ought to use more broadly applicable methodologies and cover a greater variety of merchant sectors. This study demonstrates how well HMMs predict cardholder behavior, offering merchants and stakeholders insightful information. MSc in Computer Systems 2026-03-18T00:00:00Z