GRA Projects

BACKGROUND 
The advent of scRNA-seq has had a profound impact on genomic and biomedical research. With the ability to profile gene expression at cellular resolution, we are able to significantly increase our understanding of cell types and cell states within different complex biological systems. At the same time, there are many questions that scRNA-seq technology cannot answer satisfactorily. Many biological processes, such as tissue development and cell signaling, are regulated in a spatially specific manner. To understand these processes, we need access to the spatial information of gene expression within tissues. Spatial transcriptomics is a groundbreaking molecular profiling method that provides single cell-resolution expression measurements with spatial resolution. The data generated from spatial transcriptomics experiments typically consists of gene expression levels associated with specific spatial coordinates. This spatially resolved transcriptomic data has been shown to improve our ability to identify cell types, infer cell-cell interactions, and study the organization of tissues and organs at a molecular level. The latest advance in high throughput single-cell transcriptomics is the emergence of spatial transcriptomics technologies with single-molecule resolution. Such subcellular spatial transcriptomics (henceforth “SST”) technologies provide not only the transcript abundance of each gene in a cell, but they also provide the precise subcellular locations of those transcripts, thus materializing a true spatial map of the transcriptome. Figure 1: Visualization of the graph generated for a randomly sampled cell. The nodes are colored based on the gene. 

Our lab has been working on tools for the analysis of SST (subcellular spatial transcriptomics) data and has developed the “Intracellular Spatial Transcriptomic Analysis Toolkit” (InSTAnT). This work has been described in a manuscript that is currently in revision, and I have contributed to this manuscript. InSTAnT is a toolkit that detects gene pairs and modules that co-localize within cells using specialized statistical tests. InSTAnT was used to discover several novel cell type-specific gene pair co-localizations in the brain, and its promising results demonstrate potential for future research.

Background & Question 
Advancements in DNA sequencing technologies, particularly next-generation sequencing, have accelerated the discovery of numerous genes from an extensive variety of species. The increased number of resulting protein sequences creates an opportunity to expand protein engineering, but also presents a challenge as many gene product molecular functions are poorly annotated. Protein engineering holds great promise for a wide range of human endeavors, such as the development of therapeutics drugs and gene editing, through producing protein variants that enhance the original function or are entirely novel [1]. Machine learning (ML) has been increasingly coopted for protein engineering via computational, physics based rational design. A critical component of employing ML for protein engineering is the development of a model that predicts the fitness of a protein given its sequence. This method has already yielded several algorithms that can successfully predict the effects of a mutation on function given evolutionary information of homologous sequences [2][3]. Protein language models (PLM) have been found to generate state-of-the-art representations of biological properties and achieve impressive prediction performance in protein prediction related tasks [4]. 

During Spring of 2023 I found that using a Multilayer Perceptron taking as input a PLM embedding of a mutated amino acid with a protein sequence can accurately predict protein function (Fig. 1). However, the Deep Mutational Scanning (DMS) assay data used to train my novel Multilayer Perceptron was produced in vitro and is thus subject to errors and noise as are all in vitro experiments [5]. To execute data valuation, cull harmful data points, and ultimately improve the quality of training data I will integrate a novel data quality valuation method into the previously developed language model-based deep neural network protein fitness.

INTRODUCTION 
Machine learning plays a crucial role in computational biology, enabling the efficient analysis of large datasets for generating reliable predictions [1][2][6]. The Torres Lab's SAPH-ire TFx model has shown promise in predicting the functional significance of PTMs in proteins from different eukaryotic families, potentially aiding in disease identification and treatment [1][2][3]. To enhance the robustness of PTM functionality predictions and gain insights into different prediction approaches, exploring multiple machine learning algorithms through ensemble techniques is important [6][10]. Combining diverse algorithms allows for more reliable outcomes and evaluation of each model's prediction strategies [6][10]. To address the evolving nature of biological data and prevent a decrease in accuracy caused by conceptual drift, continuous model updates are necessary [4][5] . 

As a trainee in the Torres lab, my goal is to improve the current pipeline by integrating ensemble techniques, including the utilization of different models such as One Class SVM alongside SAPHire TFx, and incorporating them with the previous cross-validation pipeline. This comprehensive approach involves cross-comparing the results between the models during the cross-validation process and retraining models if accuracy falls below a specific threshold. By implementing these advancements, I aim to develop a robust framework for analyzing and interpreting PTM functionality, gaining insights into each model's decision patterns, and ensuring more accurate predictions. These efforts will address the challenges posed by evolving PTM datasets and contribute to advancements in the field.

Background 
Since genome-wide association study (GWAS) became available, researchers have identified numerous associations between human disease and genetic variants. It allows us to understand how human genetic traits are related to complex diseases and to identify novel genes related to specific diseases [1]. However, there is a lack of diversity in human genetics studies. Many of them, especially GWAS studies, focus on European populations. Some traits share similar genetic effects on phenotypes regardless of people’s ancestral heritage. For example, skin color is highly heritable and has several replicating SNPs. Fifty-nine pigmentation-associated SNPs were identified in both Africans and Europeans [3]. On the other hand, other traits might have genetic effects on specific traits depending on ancestry groups. With these traits, the application of European-biased results to the non-European population will produce less accurate predictions [2]. In dealing with this limitation, we would like to see if there is replicating single nucleotide polymorphism (SNP) across populations and apprehend their functional and evolutionary features. Also, we would like to build the database with annotation of the variants. A machine learning method, then, will be generated to predict the likelihood of getting phenotypes when the information is given. This study can bring more insight into genetic patterns on phenotypes by ancestry group, allowing further studies to make more accurate predictions.

Background & Question 
The ocean is estimated to contain 10 to the 30 viruses(1) , making them the most abundant biological entities in aquatic ecosystems(2). Relatively little is known about marine viruses, and most viral populations discovered are novel. Virus-host interactions play a key role in ocean biogeochemistry with 20% of marine bacteria lysed daily by viruses(3). Additionally, some viruses harbor auxiliary metabolic genes that can modulate their host’s metabolism and impact carbon, nitrogen, and sulfur cycling(4). While microbial populations can impact the environment, environmental conditions also influence microbial composition. Prokaryotic genomes have been shown to be affected by environmental factors. Microbial genomes exhibit higher GC content in waters with higher amounts of phosphate and nitrate, and lower temperatures, salinity, and oxygen levels(5). Viruses are closely tied to their hosts, having shown to be synchronized with microbial processes in the surface layer of the ocean(6), but the relationship between environmental factors and viral genomic composition remains to be elucidated. In addition, viral diversity, reproductive strategies, and auxiliary metabolic genes have been shown to vary with available nutrients and depth(7).

BACKGROUND AND QUESTION 

Positive assortative mating, the tendency for individuals to choose mates that are more similar to themselves in phenotype than would be expected by chance, has been identified in humans for traits such as educational attainment, body mass index (BMI), dietary factors, and other physical measurements1,2,3. Facial phenotypic similarity between couples is another trait that has been observed across the world, and while there have been many scientific studies examining why this could be the case—reasons including implicit egotism, familiarity effect, and even game theory4,5,6—far fewer studies have aimed to quantitatively show that couples look similar to each other. This study aims to measure phenotypic facial similarity within couples through the use of 3D images acquired at Georgia Tech in The Facial Expression project.

Background 
Autoimmune conditions occur when one’s immune system starts recognizing their own body’s cells as foreign and begins attacking these healthy cells [14]. The exact reason behind autoimmune conditions is unknown, but it is thought to be a mix of both environmental and genetic factors. The genetic factors to most autoimmune conditions are poorly understood, as most autoimmune conditions are thought to be polygenic diseases, with their genetic susceptibilities being theorized to be different combinations of hundreds or thousands of alleles [2]. 

The Major Histocompatibility Complex (MHC) of the human genome, a vast region of over 8 thousand single nucleotide polymorphisms (SNPs) on chromosome 6 which encodes both cell surface proteins essential for cell recognition and the adaptive immune response in all vertebrate species, is thought to host many candidate genes for factors causing autoimmune disease [3]. Due to the polygenic nature of autoimmune conditions, many complex allelic interactions contribute to developing these conditions [3]. Machine learning techniques allow us to identify and search for nonlinear relationships among these many complex interactions in our genome [4], and these techniques can be used to robustly find different patterns in different minority cohorts. Thus, machine learning is extremely helpful in exploring the genetic underpinnings of autoimmune conditions and finding better associations between these autoimmune conditions and both genetic and environmental conditions.

BACKGROUND AND QUESTION 
Ribonucleotides are RNA precursors incorporated during replication. They are embedded in the form of rNMPs (ribonucleoside monophosphate) and are the most frequent form of DNA aberrations, as shown in Figure 1. There are approximately two misincorporated rNMPs/kb in a DNA strand (Huang, Ghosh, & Pommier, 2015). However, there are mechanisms of removing these rNMPs misincorporated in the genomic DNA. A prominent pathway is known as Ribonucleotide Excision Repair (RER) by Ribonuclease (RNase) H2 (Cornelio et al., 2017). Due to lack of RNase H2 function, rNMP incorporation increases in genomic DNA. There are few effects of rNMPs some being positive. For example, rNMPs help mating type switching in fission yeast, thus aiding reproduction during its life-cycle (Yao et al., 2013). However, there are also effects deemed harmful. In humans, a mutation in RNase H2 has been associated with Aicardi-Goutières syndrome, which is a severe childhood neuroinflammatory autoimmune disorder (Uehara et al., 2018). It is characterized by abnormal development or destruction of the white matter that affects the brain, spinal cord, and immune system.

Although there is an abundance of rNMPs in DNA, the sites of the rNMP incorporation are still poorly characterized in the human genome (Balachander et al., 2020). Last fall, my project was directed towards finding: What consensus sequences, or motifs, are around the ribonucleotides in both the wildtype (WT) and knock-out (KO) cell lines in Saccaromyces cerevisiae libraries? In addition, I found transcription factor (TF) sites near rNMPs in both the wildtype (WT) and knock-out (KO) libraries. However, the code could be improved to accommodate for more motifs among the cell lines and the position of the motifs with respect to the rNMP site. This project will help us better understand rNMP incorporation in the human genome. Once we know what motifs are present, we can better predict the locations that are more prone to rNMP incorporation and thus potentially more prone to damage.

Background 
In the last couple of decades, a reproducibility crisis has emerged in science. Many recent scientific studies are nearly impossible to reproduce, causing their results to be less credible. This reproducibility problem has extended to bioinformatics as well. Genetic ancestry refers to the population groups that their genes are derived from and is objective in nature. These population groups are inferred from the person’s sequenced genome. In this study, I will perform large-scale ancestry inference on a sample dataset and create a comprehensive guide for anybody learning the process to ensure reproducibility of the methods. 

The ultimate goal is to apply this method to several datasets which are a part of CODIGO (https://codigo.biosci.gatech.edu/), the Consortium for Genomic Diversity, Ancestry and Health in Colombia. This is a collaborative project with the National Institutes of Health (NIH) that is centered around increasing the wealth and diversity of genetic information around admixed genomes and support bioinformatics research in Colombia. The Colombian population is a result of an admixture between Europeans, Native Americans, and Africans. In addition, their ancestry is heavily dependent on where they are from in Colombia geographically. This creates a fascinating amount of diversity within the country and its population that is not yet fully understood [1-4]. 

Background and Question
Real-world Networks are complex, comprising vast webs of interconnected elements performing a diverse array of social and biological functions. Common among many networks is the pressure to be efficiently compressed either in the brain or in genetic code (1). Biological networks among molecular and cellular components are encoded at various scales in genetic material (2-5), and evolution uses these encoding to propagate network topologies in a surviving species. To answer the questions on what makes one network more compressible than another, we adapt tools from information theory to quantify the compressibility of a network.

Two ways to reduce the amount of information in the sequence is lossless and lossy compression, with lossless compression, the data originally in the file remains after compression, and all the information is restored while in our framework we will be using lossy compression, where the only the reference to recent data is saved which reduces the overall size without losing quality and is important for large scale biological networks. There are several studies done on directed networks however, few studies exist by using “real-world biological data”, therefore there is little information on complex biological networks. Some applied applications include identifying disease genes and drug targets, dysregulated pathways, and discover cell physiology in normal and disease states.

BACKGROUND AND QUESTION 
Current commercial cell growth medias are formulated to provide cell cultures with necessary nutrients. While these formulations can successfully keep cells alive in vitro, they often contain a nutrient profile that differs from that which is available to cells in the in vivo environment. The nutrient composition of commercial media, therefore, can cause “metabolic artifacts” [1] in cancer cells. These artifacts include reversing the urea cycle due to their arginine concentrations, a phenomenon which is “not observed in vivo” [1]. Plasmax is a physiological media which aims to better mimic the in vivo environment and was shown to not cause metabolic artifacts in breast cancer cell lines. The Kemp Lab is interested in verifying these results in two matched head and neck cancer cell lines that reflect radiation-sensitive and radiationresistant phenotypes. Artifacts in the metabolic profiles of cells can impact the perceived efficacy of potential treatments when tested, as the cells may respond differently than they would in the in vivo environment. If the metabolic profiles of head and neck cancer cell lines differ greatly when exposed to Plasmax as opposed to commercial media, cell culture practices should be changed accordingly to create an in vitro environment that most closely matches the in vivo environment.

BACKGROUND AND QUESTION 
Nanopore technology (ONT) is a third-generation sequencing technology becoming increasingly competitive in the sequencing market. Following Sanger and NGS sequencing methods, which rely on fragmentation and subsequent amplification to read DNA, third-generation sequencing technologies sequence a single DNA molecule directly. 

This advancement in sequencing technology comes with several key benefits. First, ONT technology can produce extremely long reads, making it a powerful tool for de novo genome assembly, structural variant identification, and traversing repetitive genomic regions. [1] Secondly, ONT has created compact, cost-effective sequencing devices such as the MinION and Flongle. These portable devices allow field-based, real-time data generation, providing valuable resources for rapid diagnostics and outbreak tracking. 

Historically, ONT has had much higher sequencing error rates than NGS sequencing (ONT, 1/50 bp-1/100 bp and Illumina, 1/1000 bp). [2,3] However, advances in equipment and software have drastically improved ONT’s error rate performance; some researchers recently claim they have achieved fidelities rivaling Illumina’s. With these advances, ONT is starting to present a viable alternative to NGS sequencing, especially in cases of de novo assembly, as well as when cost and time represent a critical factor.

BACKGROUND AND QUESTION
Mental health disorders pose a significant challenge to public health and individual well-being. Early identification and prediction of mental health outcomes can lead to targeted interventions and improved treatment outcomes. This research proposal aims to develop a predictive modeling framework for mental health outcomes, schizophrenia, bipolar disorder, anxiety disorder, and major depressive disorder.

Schizophrenia is a severe and chronic mental health disorder characterized by a range of symptoms, including hallucinations, delusions, disorganized thinking, and impaired cognitive functions. It affects approximately 1% of the global population, making it one of the most prevalent psychiatric disorders worldwide. Schizophrenia has a significant impact on the lives of individuals affected and their families. The symptoms of schizophrenia can be distressing and debilitating, leading to impairments in social and occupational functioning. The disorder often results in a decreased quality of life, increased risk of unemployment, poverty, homelessness, and increased reliance on healthcare services. Individuals with schizophrenia are at an increased risk of developing comorbidities such as substance use disorders, depression, anxiety disorders, and physical health conditions. The presence of comorbidities further complicates the management of schizophrenia and requires integrated and multidisciplinary approaches.

Background and Question
Enteric pathogens are major causes of mortality worldwide, especially in children in industrializing countries (1–3). Bacteria such as Enterococcus faecium, Escherichia coli, Salmonella enterica, and viral pathogens such as Norovirus, can be responsible for these infections. These microbes represent major global threats to human health, with E. faecium and E. coli being members of the “ESKAPEE” group (E. faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter spp., and E. coli), pathogens with high infection and mortality rates and high rates of antibiotic resistance (4). Early infection with enteric pathogens can have profound effects on child development, including both cognitive and physical, and chronic downstream effects such as diarrhea, which can significantly alter the patient’s microbiome or result in death. There have been some recent efforts to characterize the impacts of enteric pathogens, such as E. coli, on childhood infections (1,3,5), which have uncovered unexpected co-occurrence of bacterial pathogens and commensals in the gut, as well as other effects, but the routes of infection in children and the effects of specific pathogens on microbiome changes remain largely unknown. We therefore aim to characterize enteric infections in early childhood via use of metagenomics, namely the route of acquisition, and how timing of infection affects subsequent growth and development of both the child and its microbiome. Metagenomic sequencing is untargeted in this application, allowing for a comprehensive analysis of all microbes within a sample. This will allow us to not only examine specific enteric pathogens of interest, but also the overall microbiome profile of study subjects.

ABSTRACT
Prokaryotic gene prediction is largely misconceived as a solved problem in bioinformatics. There are many challenges to be solved: 1) the existing tools can hardly detect short genes (<180 nucleotides) although they are highly sensitive in finding long genes, their sensitivity decreases noticeably in finding shorter genes and 2) the false positive prediction in these gene annotation methods gives rise to hypothetical proteins, which may or may not be accurate; therefore, it is crucial to explore the reason behind these false positives. This proposal aims to investigate the potential benefits of leveraging the predicted 3D structure of potentially encoded proteins to enhance the accuracy of ProtiGeno, a transformer-based deep learning tool for prokaryotic gene prediction developed from our prior project. The objective is to acquire the 3D structure of both coding and noncoding regions in prokaryotes using state-of-the-art techniques such as AlphaFold2 and ESMFold. By doing so, we seek to explore the correlation between secondary structure patterns and their role as distinguishing features in the prediction of prokaryotic genes. This study has the potential to advance our understanding of gene prediction methodologies and contribute to the refinement of ProtiGeno's predictive capabilities.

BACKGROUND
G-protein-coupled receptors (GPCRs) are seven transmembrane receptors that couple with Gproteins on activation by external stimuli to trigger intracellular signaling cascades [1]. GPCRs comprise the largest family of membrane receptors that are prominent drug targets (35% of all approved drugs exploit GPCR signal transduction) [2]. GPCRs are thus clinically relevant targets for the development of new as well as repurposed drugs. 

This study focuses on identifying ligands for a class A cannabinoid receptor-like GPCR called GPR119 (glucose-dependent insulino tropic receptor) [3]. Pancreatic β and intestinal enteroendocrine L cells show significant levels of GPR119 expression on activation by oleoylethanolamide (OEA) [3, 4]. GPR119 participates in feeding behavior and glucose homeostasis signaling cascades via G-protein (specifically Gs) coupling, thus triggering the cyclic AMP pathway, and causing adenylate cyclase activation [4]. GPR119 activation initiates the release of incretins from the intestine and insulin from the pancreas and is also responsible for weight decline and reduced food intake in rats [4]. Determining pharmacologically relevant agonist targets for GPR119 is essential to identify lead compounds for management of metabolic disorders like type 2 diabetes and obesity [3,4]. GPR119 ligands are also determined to be a viable novel treatment for metabolic-associated fatty liver disease

Background:
Salt marshes are areas of intertidal grasslands. They are important for coast maintenance, erosion control,
carbon regulation, and they supply raw materials and food. One plant that dominates the North American
wetlands is Spartina alterniflora. This smooth cordgrass plays a large role in wetland maintenance through soil stabilization and nutrient recycling [1]. In salt marshes, there is a gradient in the phenotypic height of S.
alterniflora, wherein the plants located closer to the ocean are taller while the plants growing closer to the
shoreline are shorter [2]. Plants of this species with different height phenotypes have different microbial
populations [2, 3]. These populations are predominantly composed of sulfur-oxidizing and sulfate-reducing
bacteria [3]. The bulk soil metagenomes of bulk sediment, rhizosphere, and endosphere associated with S.
alterniflora sediment compartments of tall and short phenotypes were obtained by the Kostka lab and their
microbiome was analyzed [3].

To improve our understanding of the ecological interactions within the salt marsh associated to S. alterniflora, we are studying the viral populations found in the different sediment compartments and plant phenotypes. Viruses are the most prominent biological entities in the ocean and impact their surrounding ecosystems [4]. Through the lytic cycle, viruses infect the surrounding microbial populations, replicate within them, and then lyse the host while their progeny burst into the environment. Through this process, the viruses regulate the population densities of surrounding microbes, impacting the diversity of microbial populations [5,6]. In the lysogenic cycle, the genome of temperate phage integrates into its host’s and replicates with it, with no lysis occurring. These temperate phages can benefit their host by carrying Auxiliary Metabolic Genes (AMGs) which can play a role in boosting host metabolism [4]. With this project, we aim to characterize the viral populations associated with S. alterniflora, and the role they play in the maintenance of S. alterniflora in salt marshes. We intend to do so by studying the prevalence, taxonomy, host associations, and gene composition of viral populations in the three different sediment compartments at the different height phenotypes, and analyzing the relationship between the viruses, the microbial populations, and S. alterniflora.

Objective
To create a tool that considers clinical risk factors for colorectal cancer and polygenic risk scores (PRSs) derived from large-scale genomic data in patient risk stratification to predict the risk of individuals contracting CRC with improved discriminatory accuracy between both incident and prevalent case and control subjects. This study has three main objectives: (1) develop meta & ancestry-specific PRSs for UK Biobank and All of US cohorts, (2) create a Cox-clinical risk score for CRC using identified risk factors, (3) computing an integrated risk prediction tool (IRT) using both clinical data and genetic information to improve risk stratification and this tool would be made available for use on the R Shiny App, (4) finding the gain in accuracy of prediction when genetic data are added to clinical risk factors.

Background
Colorectal cancer (CRC) is among the most prevalent and preventable forms of cancer worldwide. There is increased awareness of a strong genetic component to CRC risk, with the identification of several high penetrance alleles that predict increased CRC susceptibility. [1] Although risk factors often influence cancer development, most do not directly cause cancer. Over the past decade, genome-wide association studies (GWASs) for sporadic CRC, which constitutes most cases, have identified ~60 association signals at over 50 loci.

Predictive genetic testing is the use of a genetic test in an asymptomatic person to predict future risk of disease. These tests represent a new and growing class of medical tests, differing fundamentally from conventional medical diagnostic tests. The hope underlying such testing is that early identification of individuals at risk of a specific condition will lead to reduced morbidity and mortality through targeted screening, surveillance, and prevention. [2]

Background and Question
Sepsis is a life-threatening condition due to the body’s extreme reaction to an infection.
It is a cascade of immunological reactions triggered by an initial infection. The primary
organs where sepsis attacks originate are the lungs, urinary tract, and gastrointestinal
tract. If sepsis is not treated within an advisable duration it can lead to tissue damage,
organ failure, and even death [1]. The number of sepsis occurrences is steadily
increasing (close to 200,000 yearly in the US). Sepsis occurs when biological/
immunological chemicals released in the bloodstream to fight an infection trigger
inflammation in neighboring tissues and organs. These cascades of changes lead to
organ dysfunction and frequently even death. In such conditions, the immune system no
longer fights against the pathogen, instead begins to turn against itself. This makes
medication extremely difficult and tricky. Septic shock is the most severe degree of
sepsis and is diagnosed when a patient’s blood pressure drops to dangerously low
levels. Sepsis also has immense economical implications in the US as well- it is the
number 1 cost of hospitalization in the US owing to the high nursing and medical cost,
and sepsis costs approximately $62 billion annually (only a small proportion) [2].

Sepsis-triggered inflammation is one of the leading causes of death among patients
admitted to an ICU rather than the primary infection.

Acute Respiratory Distress Syndrome (ARDS/ ARS) is a devastating complication of
severe sepsis. Both Sepsis and ARDS fundamentally have the same underlying
characteristics- inflammation and endothelial dysfunction. Patients diagnosed with
sepsis-induced ARDS have higher fatality and lower survival rates. [3]

BACKGROUND AND QUESTION
The Lachance Lab works with novel Prostate Cancer (PCa) Biological Datasets of men of African Ancestry (AA). These data sets have been largely made available by Men of African Descent and Carcinoma of Prostate (MADCaP) consortium. Our aim is to understand and further study the high prevalence of PCa in men of African Ancestry. Despite men of African Ancestry (AA) having the highest mortality rate from PCa, we have little knowledge about the variants associated with it. Preliminary research conducted in the lab has discovered novel GWAS (Genome-Wide Association Studies) hits. Additionally, considerable heterogeneity was discovered in the genetic architecture of PCa within West, East and South Africa.

To gain a deeper insight into the heterogeneity observed with our Genome Wide Association Studies (GWAS) results, Linkage Disequilibrium (LD) is crucial as it allows the identification of genetic markers that tag the causal variants [1]. LD is the difference between frequency of a particular combination of alleles at two loci that is observed and the expected frequency for random association. In a population where there is no mutation, selection of specific gene combinations will result in LD. However, genetic recombination breaks this LD [2].

Genetic recombination is an important process that gives rise to novel allele combinations enabling evolution in species [3]. Recombination rates are highly variable across species and populations [4]. The recombination rate is represented as the ratio of genetic distance and physical distance that forms a recombination map. Genetic recombination maps are essential for the evolutionary analysis [5] of populations, one important application being the estimation of allele ages. Presently, there are only a limited number of recombination maps available for the sub-Saharan African population for West Africa [6] and the Khoe San population of South Africa [5]. Therefore, we need accurate genetic recombination maps to study the heterogeneity in the African populations.

Previously, I found various tools that are publicly available which could be used to generate accurate and reliable genetic recombination maps. For initial trial, I used HapMap [7] and Thousand Genomes Project [6] data as input for these tools. I aim to find the best tools, most-suited to create recombination maps with MADCaP Datasets.
 

BACKGROUND
The incidence of prostate cancer (CaP) varies among different populations. Individuals of African American ancestry face a 75% higher incidence risk than non-Hispanic Whites1. Among males in African populations, CaP ranks as the leading cause of cancer-related deaths. The factors contributing to this disparity involve a combination of genetic and non-genetic factors, many of which are not yet fully understood2. Conducting genetic studies in Africa presents challenges due to limited infrastructure and resources, leading to lower diagnosis rates. Additionally, variations in medical care access, risk assessment, and lifestyle factors within the African continent emphasize the need for subpopulation studies3.

Given the disease’s geographical and ethnic diversity and its higher mortality rate compared to other populations, it is crucial to identify specific characteristics unique to certain regions, which may not be apparent in pan-cancer studies. However, assessing the genetic burden of CaP in Africa is challenging due to limited data availability and inaccurate cancer incidence measures.

Introduction:
Crohn’s disease (CD) is an inflammatory bowel disease that affects the gastrointestinal tract. Most patients with Crohn’s disease take drugs such as anti-TNF agents and thiopurines to alleviate their symptoms. Patients who are less responsive to their initial treatment or with bowel obstruction may undergo surgery with careful medical assessment. However, surgery does not provide a cure for Crohn’s Disease and the risk of postoperative recurrence is 44-55% after 10 years1.

Aberrant RNA splicing contributes to various diseases such as spinal muscular atrophy, cancers, and Hutchinson–Gilford Progeria Syndrome2,3. The genetic variants affecting the RNA splicing event is called splicing quantitative loci (sQTL). Recent studies predicted that sQTL correlates with certain diseases and might be a major contributor to diseases like Parkinson’s disease4,5. Previous research from Gibson lab shows that CD individuals with distinct transcript profile and altered exon usage might be caused by altered splicing events, which is also called spliceopathy6.
In this project, our aim is to investigate the splicing events in post-operative patients. By comparing the splicing profile of recurrence patients with non-recurrence patients, we intend to find differential isoform usage or differential exon usage that is associated with CD recurrence. Combining alternative splicing analysis, splicing QTL, and genome-wide association study (GWAS), I will address the following questions:

1) Does alternative splicing contribute to the risk of postoperative recurrence?
2) Do male and female show the same pattern of alternative splicing in CD recurrence?
3) Which genetic loci control the differential splicing between the recurrence and non-recurrence patients?

The main objective of this project is to predict the recurrence of CD from genotype and expression data. The integration of sQTL into GWAS and RNA splicing analyses provides insights into the mechanism behind post-operative disease recurrence.