E. Adetiba

Ensembling of EGFR Mutations' based Artificial Neural Networks for Improved Diagnosis of Non-Small Cell Lung Cancer

In this research work, we built and ensembled different EGFR microdeletion mutations’ based Artificial Neural Networks(ANNs) for improved diagnosis of Non-Small Cell Lung Cancer(NSCLC). We developed two novel algorithms, namely; Genomic Nucleotide Encoding & Normalization (GNEN) algorithm to encode and normalize the EGFR nucleotides and SimMicrodel algorithm to programmatically simulate microdeletion mutations. Sample patients’ data with microdeletion mutations were extracted from online EGFR mutation databases and the two novel algorithms (implemented in MATLAB) were applied to these data to generate appropriate data sets for training and testing of the networks. The networks after proper training, were combined using minimum error voting ensembling to predict the number of nucleotide deletions in NSCLC patients. Using this ensembling approach, our simulations achieved predictions with minimal error and provides a basis for diagnosing NSCLC patients using genomics based ANN. Key Words: ANN, EGFR, GNEN, NSCLC, LM

Experimentation using short-term spectral features for secure mobile internet voting authentication

We propose a secure mobile Internet voting architecture based on the Sensus reference architecture and report the experiments carried out using short-term spectral features for realizing the voice biometric based authentication module of the architecture being proposed. The short-term spectral features investigated are Mel-Frequency Cepstral Coefficients (MFCCs), Mel-Frequency Discrete Wavelet Coefficients (MFDWC), Linear Predictive Cepstral Coefficients (LPCC), and Spectral Histogram of Oriented Gradients (SHOGs). The MFCC, MFDWC, and LPCC usually have higher dimensions that oftentimes lead to high computational complexity of the pattern matching algorithms in automatic speaker recognition systems. In this study, higher dimensions of each of the short-term features were reduced to an 81-element feature vector per Speaker using Histogram of Oriented Gradients (HOG) algorithm while neural network ensemble was utilized as the pattern matching algorithm. Out of the four short-term spectral features investigated, the LPCC-HOG gave the best statistical results with statistic of 0.9127 and mean square error of 0.0407. These compact LPCC-HOG features are highly promising for implementing the authentication module of the secure mobile Internet voting architecture we are proposing in this paper.

Implementation of XpertMalTyph: An Expert System for Medical Diagnosis of the Complications of Malaria and Typhoid

The dearth of medical experts in the developing world has subjected a large percentage of its populace to preventable ailments and deaths. Also, because of the predominant rural communities, the few medical experts that are available always opt for practice in the few urban cities. This consequently puts the rural communities at a disadvantage with respect to access to quality health care services. In this work, we designed and implemented XpertMalTyph; a novel medical diagnostic expert system for the various kinds of malaria and typhoid complications. A medical diagnostic expert system uses computer(s) to simulate medical doctor skills in diagnosis of ailments and prescription of treatments, hence can be used to provide the same service in the absence of the experts. XpertMalTyph is based on JESS (Java Expert System Shell) programming because of its robust inference engine and rules for implementing expert systems.

Experimental Investigation of Frequency Chaos Game Representation for in Silico and Accurate Classification of Viral Pathogens from Genomic Sequences

This paper presents an experimental investigation to determine the efficacy and the appropriate order of Frequency Chaos Game Representation (FCGR) for accurate and in silico classification of pathogenic viruses. For this study, we curated genomic sequences of selected viral pathogens from the virus pathogen database and analysis resource corpus. The viral genomes were encoded using the first to seventh order FCGRs so as to produce training and testing genomic data features. Thereafter, four different kernels of naive Bayes classifier were experimentally trained and tested with the generated FCGR genomic features. The performance result with the highest average classification accuracy of 98% was returned by the third and fourth order FCGRs. However, due to consideration for memory utilization, computational efficiency vis-a-vis classification accuracy, the third order FCGR is deemed suitable for accurate classification of viral pathogens from genome sequences. This provides a promising foundation for developing genomic based diagnostic toolkit that could be used to promptly address the global incidence of epidemics from pathogenic viruses.

Breathogenomics: A Computational Architecture for Screening, Early Diagnosis and Genotyping of Lung Cancer

The genome sequences of some genes have been implicated to carry various mutations that lead to the initiation and advancement of lung cancer. In addition, it has been scientifically established that anytime we breathe out, chemicals called Volatile Organic Compounds (VOCs) are released from the breath. Hundreds of such VOCs have been uniquely identified from samples of breathe collected from lung cancer patients, which make them viable as chemical biomarkers for lung cancer. Based on the foregoing scientific breakthroughs, we developed breathogenomics, a computational architecture for screening, early diagnosis and genotyping of lung cancer victims anchored on the analysis of exhaled breath and mutational profiles of genomic biomarkers. The architecture contains two important sub-modules. At the first sub-module, the exhaled breadths of smokers or persons that are at risk of lung cancer are collected and appropriate computational algorithms are employed to determine the presence of any of the VOC biomarkers. Next, a patient with any VOC biomarker in the exhaled breath proceeds to the second sub-module, which contains appropriate computational models for the detection of mutated genes. Once mutations are detected in any of the biomarker genes found in a given patient, such patient is recommended for targeted therapy to promptly curtail the progression of the mutations to advanced stages. The breathogenomics architecture serves as a generic template for the development of clinical equipment for breath and genomic based screening, early diagnosis and genotyping of lung cancer. In this paper, we report the preliminary result obtained from the prototype that we are currently developing based on the architecture. Constructing a lung cancer early diagnosis/screening system based on the prototype when fully developed will hopefully minimize the current spate of deaths as a result of late diagnosis of the disease.

In Silico Prediction of 3D Structure of Anopheles Gambiae ABCC12 Protein

In this paper, the Anopheles gambiae ABCC12 MRP protein domain sequence which contained 216 residues was obtained from the NCBI database in its fasta format (NCBI entry EAA12438.4). This MRP protein sequence was Gapped Blast using BLOSUM 62 matrix with an E-value cut-off of 0.000001 to identify the closest homologous structure as at the date of the study (Nov 2016). Additionally, the sequence was aligned with three prediction modelers, which are the Modeller v9.15 alignment script, the Swiss Model Server and the Raptor-X server for modelling based on the server’s automated choice for a suitable template. The structure predicted by Raptor-X has a higher percentage (90.1%) of residues in the most favored regions as compared to Modeller and Swiss-Model (86.5%). This paper further unveils the quality of structure predicted during homology modeling and the diverse correlation as well as the significance of ABCC12 in drug design for malaria vector.