Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Polygenic risk scores for prediction of tuberculosis susceptibility in an admixed South African population
Thesis (MSc)--Stellenbosch University, 2022.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
Stellenbosch : Stellenbosch University
2022
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613764050747392 |
|---|---|
| access_status_str | Open Access |
| author | Ndong Sima, Carene Anne Alene |
| author2 | Moller, Marlo |
| author_browse | Moller, Marlo Ndong Sima, Carene Anne Alene |
| author_facet | Moller, Marlo Ndong Sima, Carene Anne Alene |
| author_sort | Ndong Sima, Carene Anne Alene |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MSc)--Stellenbosch University, 2022. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/125079 |
| institution | Stellenbosch University (South Africa) |
| language | English |
| last_indexed | 2026-06-10T12:41:19.685Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2022 |
| publishDateRange | 2022 |
| publishDateSort | 2022 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/125079 Polygenic risk scores for prediction of tuberculosis susceptibility in an admixed South African population Ndong Sima, Carene Anne Alene Moller, Marlo Uren, Caitlin Schurz, Haiko Chimusa, Emile Stellenbosch University. Faculty of Medicine and Health Sciences. Dept. of Biomedical Sciences. Molecular Biology and Human Genetics. Tuberculosis -- Susceptibility -- South Africa Human genetics -- South Africa Polygenic risk score UCTD Thesis (MSc)--Stellenbosch University, 2022. ENGLISH ABSTRACT: Progression from latent to active TB has been shown to be influenced by host genetic factors, which prompted genome-wide association studies (GWAS) that explored the genetic susceptibility to TB. Although single nucleotide polymorphisms (SNPs) have been associated with TB susceptibility, they all exhibit small effect sizes which individually contribute little to disease risk. Therefore, suggesting that disease outcome is influenced by the additive effect of multiple SNPs. To exploit these findings, polygenic risk scores (PRS) have been proposed to assess an individual’s genetic liability to a trait. However, similar to GWAS, PRS have mostly been assessed not only in European populations but also mostly for non-communicable diseases and traits. Although efforts in the area are expanding to diverse populations and two-way admixed individuals, little is known about how to effectively construct PRS in more genetically complex populations as well as for infectious diseases. Here, we explored different approaches for constructing PRS for prediction of TB susceptibility in a multi-way admixed South African (SAC) population, using GWAS summary statistics from a medium-sized Ghanaian (n=3 298) and a small independent SAC population (n=787). We assessed and compared the predictive performance of three PRS construction tools, namely PRSice2 (standard approach), LDpred2 (Bayesian approach), and Lassosum2 (LASSO penalized regression). We also evaluated the predictive power of different combinations of TB-PRS models and nongenetic factors. Our results showed that LDpred2 consistently yielded better prediction accuracy compared to the non-Bayesian PRS models, with a mean AUC of 55.2% compared to 52.1% for Lassosum2 and 53.6% for PRSice2 with the Ghanaian selected variants; and a mean AUC of 55.3% compared to 50.2% and 43.9% with the SAC selected variants, for Lassosum2 and PRSice2 respectively. Notably, we observed a 18% relative loss of power for PRSice2 in the SAC-PRS model compared to its Ghanaian counterpart. Additionally, we showed that the linear mixture of PRSs outperforms models that use a single discovery population. We attained 22% relative improvement in prediction accuracy with the multi-ethnic model compared with the SAC-only PRS model and a modest improvement (2%) compared with Ghanaian-only PRS model. Furthermore, we demonstrated the strong association of age with TB risk and showed that the combination of genetic and nongenetic risk factors showed the most predictive power gain (AUC [95%CI]=0.619 [0.543 – 0.693]). Our study showed that the choice of the PRS tool used can greatly influence the predictive accuracy of model in an admixed population. Additionally, we highlight that in the event that nongenetic factors are strongly associated with disease outcome, their inclusion in a linear mixture of risk models can result in a substantial gain in prediction power. AFRIKAANSE OPSOMMING: Daar is getoon dat progressie van latente na aktiewe TB beïnvloed word deur gasheergenetika. Hierdie bevinding het gelei tot die uitvoer van genoomwye assosiasiestudies (GWAS) wat die genetiese vatbaarheid vir TB ondersoek het. Alhoewel verskeie enkelnukleotied polimorfismes (ENPs) met TB geassosieer is, het die meeste hiervan klein effekte wat individueel min bydra tot siekterisiko. Dit dui daarop dat siekte-uitkoms beïnvloed word deur die bykomende effekte van verskeie ENP’s. Om hierdie bevindinge te ontgin, is poligeniese risikotelling (PRT) voorgestel om ‘n individu se genetiese aanspreeklikheid vir ‘n eienskap te bepaal. Net soos vir GWAS, is PRT nie net meestal in Europese bevolkings geassesseer nie, maar grootliks ook net vir nie-oordraagbare siektes en eienskappe. Alhoewel studies in die gebied uitbrei na diverse en tweerigting-gemengde bevolkings, is min bekend oor hoe om PRT effektief te skep in bevolkings met komplekse genetiese vermenging, asook vir aansteeklike siektes. Hier het ons verskillende benaderings ondersoek om PRT saam te stel vir die voorspelling van TB vatbaarheid in ‘n meervoudige gemengde Suid-Afrikaanse (SAK) bevolking, deur gebruik te maak van GWAS-opsommingstatistieke van ‘n mediumgrootte Ghanese (n=3 298) en ‘n klein onafhanklike SAK-bevolking (n=787). Ons het die voorspellingsvermoë van drie PRSkonstruksie-instrumente geassesseer en vergelyk, naamlik PRSice2 (standaardbenadering), Ldpred2 (Bayesiaanse benadering) en Lassosum2 (LASSO gepenaliseerde regressie). Ons het ook verskillende kombinasies van TB-PRT-modelle en nie-genetiese faktore geëvalueer om te bepaal hoe dit voorspelling beïnvloed. Die resultate het getoon dat Ldpred2 konsekwent beter voorspellingsakkuraatheid gelewer het in vergelyking met die nie-Bayesiaanse PRS-modelle, met ‘n gemiddelde AUC van 55.2% in vergelyking met 52.1% vir Lassosum2 en 53.6 vir PRSice2 met die Ghanese geselekteerde variante; en ‘n gemiddelde AUC van 55.3% in vergelyking met 50.2% en 43.9% met die SACKgeselekteerde variante, vir Lassosum2 en PRSice2 onderskeidelik. Ons het veral ‘n verlies van onderskeidingsvermoë van 18% vir PRSice2 in die SAK PRT-model waargeneem in vergelyking met die Ghanese eweknie. Daarbenewens het ons getoon dat die lineêre mengsel van PRT beter presteer as modelle wat ‘n enkele ontdekkingspopulasie gebruik. Ons het ‘n 22% relatiewe verbetering in voorspelling akkuraatheid behaal met die multi-etniese model in vergelyking met die SAK-net PRT model en ‘n beskeie verbetering (2%) in vergelyking met Ghanese-alleen PRT model. Verder het ons die sterk assosiasie van ouderdom met TB-risiko getoon en gewys dat die kombinasie van genetiese en nie-genetiese risikofaktore die mees voorspellende kragtoename getoon het (AUC [95%CI]=0.619 [0.543 – 0.693]). Ons studie het getoon dat die keuse van die PRT-instrument wat gebruik word, die voorspellingsakkuraatheid van die model in ‘n gemengde populasie Stellenbosch University https://scholar.sun.ac.za V grootliks kan beïnvloed. Daarbenewens beklemtoon ons dat indien nie-genetiese faktore sterk geassosieer word met siekte-uitkoms, die insluiting daarvan in ‘n lineêre mengsel van risikomodelle kan lei tot ‘n aansienlike toename in voorspellingsvermoë. Masters 2022-02-23T19:06:00Z 2022-04-29T12:53:03Z 2022-02-23T19:06:00Z 2022-02 Thesis http://hdl.handle.net/10019.1/125079 en Stellenbosch University x, 75 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Tuberculosis -- Susceptibility -- South Africa Human genetics -- South Africa Polygenic risk score UCTD Ndong Sima, Carene Anne Alene Polygenic risk scores for prediction of tuberculosis susceptibility in an admixed South African population |
| title | Polygenic risk scores for prediction of tuberculosis susceptibility in an admixed South African population |
| title_full | Polygenic risk scores for prediction of tuberculosis susceptibility in an admixed South African population |
| title_fullStr | Polygenic risk scores for prediction of tuberculosis susceptibility in an admixed South African population |
| title_full_unstemmed | Polygenic risk scores for prediction of tuberculosis susceptibility in an admixed South African population |
| title_short | Polygenic risk scores for prediction of tuberculosis susceptibility in an admixed South African population |
| title_sort | polygenic risk scores for prediction of tuberculosis susceptibility in an admixed south african population |
| topic | Tuberculosis -- Susceptibility -- South Africa Human genetics -- South Africa Polygenic risk score UCTD |
| url | http://hdl.handle.net/10019.1/125079 |
| work_keys_str_mv | AT ndongsimacareneannealene polygenicriskscoresforpredictionoftuberculosissusceptibilityinanadmixedsouthafricanpopulation |