MINISTRY OF EDUCATION AND TRAINING NONG LAM UNIVERSITY HO CHI MINH CITY FACULTY OF BIOLOGICAL SCIENCES DEVELOPMENT OF DNA METHYLATION BIOMARKERS FOR THYROID CYTOLOGY DIAGNOSIS Major : BIOTECHNOLOGY Student name :HO THI KIM CUONG Student ID : 18126017 Academic year : 2018 - 2023 Thu Duc City, 03/2023 MINISTRY OF EDUCATION AND TRAINING NONG LAM UNIVERSITY HO CHI MINH CITY FACULTY OF BIOLOGICAL SCIENCES GRADUATION THESIS DEVELOPMENT OF DNA METHYLATION BIOMARKERS FOR THYROID CYTOLOGY DIAGNOSIS Advisors Student Luu Phuc Loi, Ph.D Ho Thi Kim Cuong Nguyen Vu Phong, Ph.D 18126017 ACKNOWLEDGEMENTS A completed study would not be done without any assistance. Therefore I would like to thank my supervisors Luu Phuc Loi, Ph., and Nguyen Vu Phong, Associate Professor Ph. for all their help and advice with this thesis. Their immense knowledge and plentiful experience have encouraged me in all the time of my academic research and daily life.
And I would like to thank all the members of the VnPathoinformatics group for their support and suggestions to successfully complete this thesis, especially Nguyen Phan Xuan Truong, MD, Msc and Le Nhat Thong. Finally, my appreciation also goes out to my family and friends for their encouragement and support all through my studies. CONEIRMATION AND COMMITMENT My name is Ho Thi Kim Cuong student ID: 18126017, class: DH18SHD (Phone: 0376255781, email: 18126017@st.vn), of Biotechnology, Nong Lam University Ho Chi Minh city. This is my graduate thesis, which I researched and wrote.
The information and data are all entirely accurate and honest. This thesis's research was not previously submitted to another higher education institution for a degree or diploma. Thu Duc City, March 17, 2023 Student’s signature Ho Thi Kim Cuong ABSTRACT In cytology, indeterminate thyroid nodules (ITNs) represent a challenge for both physicians and pathologists. ITNs include atypia of undetermined significance/ follicular lesion of undetermined significance, follicular neoplasm/suspicious for a follicular neoplasm, and suspicious for malignancy.
Molecular testing for ITNs provides the additional pre-operative classification of cancer risk but has a low positive predictive value (PPV = 47%). These facts suggest that there is a need for a highly accurate thyroid nodule diagnostic test. This study aims to employ a machine learning algorithm to predict the malignancy of ITNs by profiling DNA Methylation. Data was collected in the GEO database with 187 reduced representation bisulfite sequencing thyroid tissues.
We utilized this data, called differentially methylated cytosines (DMC) and called differentially methylated region (DMR) to find differential signature DNA methylation in non-malignant (n) and malignant nodules (m). Next, we used specific differential signature DNA methylation profiling to build the model. Using a machine learning technique, a diagnostic model was developed using training data containing 150 samples (m=64; n=86) and testing data containing 37 samples (m=16; n=21). Lastly, the association between markers and associated genes was explored.
As result, We found 135 significant DMCs and 306 significant DMRs for malignant and non- malignant thyroid tissues. After that, we reduced features to find specific differential DNA methylation but still make sure the model efficiency was well-behaved by using the best subset selection method. As the result, we constructed the RandomForest model for diagnosis using 21 features. Test data were used to determine the model's effectiveness parameters (AUC = 0.
Finally, we found thyroid cancer is intimately associated with the immune system. This data showed that epigenetic testing could provide outstanding diagnostic accuracy for thyroid nodules by evaluating tissue-specific DNA methylation. This study will be validated with ITNs cytology samples and developed into a high- performance diagnostic tool in the future Keywords: indeterminate thyroid nodule, fine-needle aspiration (FNA), DNA methylation, Random Forest algorithm. TABLE OF CONTENTS Page ACKNOWLEDGEMENTStiisiscsiccsssseintanncprarainrecoancinneieantinatinarea anne i CONFIRMATION AND COMMITMENT.cssssssssescsssssssssssesssssseessssesessssesessueesesuseessaneesssneersnneessnnees ii STRAT seasenseesoessarsesconsssssesseusanssesonssaseeceousonseettonseseeesonssarestousaseensounesenettousesssstsusssuseesonnasaesttouasaceeseuvasanesseuucs iii TABLE OF CONTENTS giscnisisiseeccasincsnsnvninancostnainiean nian ainsi iv LIST OF ABBREVIATIONS IIRBSN9)0.
ii TIẾT OE TTGTUIK SG: oscnninbndidiadbolesaeseeniebissgbdinggisingitbiesaGaiutiiissnsglxnsatrelsiftesogignsauispasd viii CHAPTER. TT R DU LI Cu nngtrnrinstrsatosaodiiningoiiiiiitogitiaagttitbšigriitlslsgianliigshitiiqugiigasigissegad 1 1.----s-+--x+++xkrE+ H1 HH HH HH HH HH HH HH H11 11g11. 1 12200 I CCE VCS inirbniodidindbogntitaslibsisininhsixpidndggtiigi th tuiổngBãggghugi4dghatuggdisktarssiGiuRaptdtlrlStigpiliesicssesiiobsvsl 2 1:13: COHEETTES Srhigrrtttrrtotitiztfttiiittaligutttrpgttagtditsotttsrfiigiiraitiölgtsqsatltiltsgiaagsrabbgitsiiagi@apssna Z CHAPTER 2.eessssssssstsssssstesssssessessseeecsneesssneeessnseeessnneesssnseessnneeeenneeeeenseessans 3 ZoLs THYPOL CANCER: Ga. ớaớ ốc ốc CC r0 3 2:11» ]-HWEỐTQ fGGHSS sintsersrirtotistbisiilosoGirustidgrtiigiofntgtftsidititiptlfftsrifaiGiiBigatsSosxtgitttltlgsisuasnatosssbi 3 2.
Diagnosis of thyroid tumot. Fine needle aspiration (FNA).sssssssesssesssessssesssecsssessssesssessssesssseesseesssseesseessssessssessneesneenss 4 2:2, Overview DNA TTiELHVIđLOHuusiesosetusiintrbdddigutagihsatitqssgtts8tiiiotBiittigsoitiojsqiseeaqusad 6 2. DNA MHethylavOn in Can COP TT TT TỶ ca ca 8 CHAPTER 3. MATERIALS AND METHODS.
Time and location Bi 2 MACOV ial S scssisscsvsecssscvoseecousgecnszceusasnosssrnaavevecennscaneaserensueansaninoxeecenserisanessaneusuncosatesuurenenrerasnerevantweanenvenenvanatans S2: 1á Datel xaaggonggoiSaotttsetidtisnititoitbSilGtuEitltagsiidlettq8ÄtödSitnisiasiidtHồittsililallaiibiiapfsidi4biisat0tssiiagfaga 10 l3. MOTO aseenesnnnesdtiieididiDiDigiTDiSE00t8AEDiSANG810H1Gãthh5HEEIGGH11G818110381390001G11300/318810100811901415400100501808008E1. Call differentially methylated cytosines (DMCS). Differentially methylated Cytosines coscssensen go dáng gi Hà H14 dáng giáng 11g tàng oang 11 313.102, FEAtUTE SEIS CU ON aissiscincnnnneniainnmomem ERED 12 3.1: Packages MUVR It Ñ:-sezccuseseeasEnirtnieoniioisniidtirnEitilinREikidHL.2: Packages random F Orest I.
Creating a model classifier for malignant (m) and non-malignant (benign and iv normal adjacent) (n) with the Random Forest algorithim. Validation cohort in public datasets sscssiscssscvsntescsansceninaninnemnmncnsennaaenin 17 SE //16i2j0121218.21100181411112415211-21)/)L-10777 TỶ ỗỶ xxx.DiffeFenuallÿ Methylated GVLOCINES sissies 17 3. Differentially methylated regions (DMRS).e«cscceeeeriiirrrriiirriiiirrire 18 CHAPTER 4, RESULTS AND DISCUSSION sscississsssnssssussninusnnennnreeansennmneinaniemannnenien 20 A] RESUS cixasesenieesanedtoiiniisaintdtsingiitiitaSinhA880013800180301303111688044G150NGD1HH1. Profiling the specific differential DNA methylation of malignant nodules (m) and nonmalignant nodules (1) ssc 20 4.
Build the machine-learning model for discriminating thyroid tissues. Validation the model sáasssssnnsansinnnisnniasnidnieinntsirtdisagSlltSt830x81645161148881084014018110611441880018008138408178838 25 4. Gene Set analysis and the pathway of these DMC affect the development of KH EGIO-THHHDES cccssnssnnmnnemnenvemecannmane me 25 3:1:4. Genesset ana ly si sanssesaenangiginiittdtuiodiitdiitiiilltiitiiabioiiftdfiWSiIciii0GSHãĐi160380886300601381114gữ0i0u83800v08g000 25 4.
Pathway affects the development of thyroid tumOFS.-------<ccc-cceeecer 27 EZAUD VS CUS S10 EXstssysgianitiikgii3i8078i3035g7aqi8878GRagĂENG4801ã88iaauiãNtiÄaiskissgiSigiauStãmgGtadãiquiQNiadautiãAgiãi8S8j8g0iãmdiSNgtdkiastijiqRsaS8 28 CHAPTER 5. CONCLUSIONS AND SUGGESTIONS. cceeeierirerrrirer 31 Bel «CONCLUSIONS name namsanensncmmnenaaenenEn RETR 31 5.rssercosescrseeessnsecesseseseensanersaneessseesstensanseenseesonueresnessanteraneeraneecanseesuerssserssaneersuteroneesenseeaneensaneerans 31 REBEREN CES creneeeariirtrgarirsasrnndtidtrnidiriatdintrllirrftiigH48813NggihdErtagồiitzfgpgntamfixirgaaãnEHSinssitrfSnispgcilqE0 SZ LIST OF ABBREVIATIONS bp Base Pair AUC Area Under the Curve DMCs Differentially Methylated Cytosine DMLs Differentially Methylated CpGs /Loci DMRs Differentially Methylated Region DNA Deoxyribonucleic acid DNMT DNA Methyl Transferases FDR False Discovery Rate FNA Fine-needle Aspiration GEO Gene Expression Omnibus ITNs Indeterminate Thyroid Nodules NPV Negative Predictive Value PCA Principal Component Analysis PPV Positive Predictive Value RF Random Forest RRBS Reduced-Representation Bisulfite Sequencing TSH Thyroid-Stimulating Hormone vi LIST OF TABLES Page Table 2. The Bethesda System for Reporting Thyroid Cytopathology.
Patient data for the training and testing cohOoTfS. The feature selection method ccsssxccsserassessevenseaentsnemremmavenennucenavesens 21 Table 4. Confusion matrix in test data.ccececceeceesceeceesececeeseeseeeseeseeeseeseenseeneeees 23 vii LIST OF FIGURES Page Hi0HFE 2.1 FINA: Cytolosy OF DGHHGTT se seccsessassasacannensennnesnnessannaoss ĐRHESSSGHEHSGS4GIGEESSSGHS2R954E8 5 Figure 2. FNA- cytology of TINS ca scceeecsssets21662 181101610530 016615860L08g68016410/310400408e 6 Figure 2.
FNA cytology of maÌignanCy.- -- -¿ c sc+ + + + *+svrererrrrrsrrrrrkrrkrrke 6 Figure 2. Normal and cancer genomes exhibit distinct DNA methylation profiles 8 Figure 3. Detailed workflow in the studyy. Working principle of MUƯVĨ.- -- 5+ cSc+sxssrserrrrsrrrrrrrrrrrrrrrrke 12 Figure 3.
Random forest schematic .ce:cesceeceeseeeceeceeceeseeseeseeaceeceeneeneesneess 14 Figure 3. Area under ROC curve (AUROC). Components of the confusion ImAfTIX. ---- 5 5-ss+<s++xc+s+xeezeeexerrees 16 Figure 3.
Schematic of CpG amnotations.cccecceeceeseeeceeseeeceeseeseeeseeseenseeaeenes 18 ETơure 4.1, Graphical ASA tiusessobsotiitiecsbiiEBiGSES4EE6L25SXSEGDILESSEENSLAASS. The specific differential DNA methylation of malignant nodules and non-malignant nodules 21 Figure 4. Annotation of 135 DMCs with differential DNA methylation. DNA methylation levels of malignant and non-malignant with 23 sigmificant DMCS.
Boxplots showing significant differences in the DNA methylation levels abG outot 2), DMCs ita the tO El srsccosc:sntqsunaesecnvaapesitenceestineerconsnencenvaeuzannsteszeesdzeneeontite 23 Figure 4. Model Random Forest classifier for malignant and non-malignant (benign and normal adjacent) 0007277. Boxplot showing significant differences in the methylation levels at 1 out of 2] DMGsn the model, secc:ssscsnsasessnenen naman nana 25 Figure 4. Chromosome ideograms of the 21 significant DMCs in the model.
Bulk tissue gene expression for PLXNB2. Plotting a DMR in the genorme. Pathway significance from DMRS.::ceccesceeseeseeseeeseeeenseeseenees 28 viii CHAPTER 1. Problem statement Thyroid cancer, which has a relatively low disease-specific death rate, is ranked as the tenth most common cancer worldwide and the most common endocrine cancer in the World Health Organization's 2020 statistics.
In Vietnam, the number of new cases is over 5 thousand, and the number of deaths is 642 (Sung et al. Fine Needle Aspiration (FNA) 1s one of the methods used to diagnose thyroid cancer in patients with suspected thyroid nodules. The Bethesda System for Reporting Thyroid Cytopathology has become a powerful tool for directing the clinical management of thyroid fine-needle aspiration (FNA) specimens (Ali et al. Thyroid cancers are categorized into six different categories: non-diagnosis, benign follicular nodules, atypia of unknown significance/follicular lesions of unknown significance (AUS/FLUS), follicular neoplasm or suspicion of a follicular neoplasm, suspicion of malignancy, and malignancy.
On cytology, however, indeterminate thyroid nodules (ITNs) pose a problem for both doctors and pathologists. ITNs include atypia of undetermined significance/ follicular lesion of undetermined significance, a follicular neoplasm/ suspicious for a follicular neoplasm, and suspicious for malignancy (Haugen et al. The uncertainty of the risk of malignancy (10-75%) in these ITNs complicates management (Ali et al. Patients with ITNs may receive diagnostic surgery, which poses a risk of operative complications and unnecessary costs to the patient.
In around 75% ofcases, a benign thyroid nodule is discovered as the final result (Bongiovanni et al. The data for this project were used from a published study in 2020. However, to predict thyroid cancer tissue, 373 different methylated gene sites are required. So I'm thinking of new with a prediction approach that integrates really well and requires few features.
It will reduce the cost even further. Molecular testing for ITNs provides an additional pre-operative classification of cancer risk but has a low positive predictive value (PPV = 47%) (Alexander et al. These facts suggest that there is a need for a highly accurate thyroid nodule diagnostic test, specifically based on DNA methylation epigenetics. 1, DNA methylation is an important regulator of gene transcription.
Changing the gene expressions of the cells leads to inappropriate silencing of tumor suppressors or the expression of oncogenes. In normal human cells, the promoter gene is usually unmethylated and expresses the tumor suppressor gene. And tumor-suppressor genes are made silent by DNA methylation in CpG islands at the promoter because it prevents transcription factors from binding, thereby inhibiting gene expression. Inversely, oncogene promoter hypomethylation resulted in transcriptional activation.
Therefore, this study aims to employ a machine learning algorithm to predict the malignancy of ITNs by profiling DNA methylation. Objectives Identifying thyroid tumor-specific differentially methylated sites and building a model for diagnosis of malignant thyroid cancer. Contents To achieve the mentioned objective, this research conducted 4 contents: e Content 1: Call differentially methylated cytosine (DMC) e Content 2: Developed the machine learning model using the Random Forest algorithm e Content 3: Validation model using GEO database e Content 4: Call differential methylation region (DMR) and analysis pathway CHAPTER 2. LITERATURE REVIEW Literature review elucidated issues related to thyroid cancer and DNA methylation, including diagnosis, the FNA method, and DNA methylation in cancer.
Thyroid cancer The thyroid is a major endocrine gland that is situated anterior to the trachea at the base of the throat. It is composed of two wing-shaped lobes and an isthmus that connects them. Normally, during a physical examination, it is impossible to feel through the skin (Nguyen et al. Thyroid cancer, which has a relatively low disease-specific death rate, is ranked as the tenth most common cancer worldwide and the most common endocrine cancer in the World Health Organization's 2020 statistics (Sung et al.
There are four main types of thyroid cancer. These are papillary, follicular, medullary, and anaplastic. Papillary is the most common type. Thyroid nodules In the general population, thyroid nodules are relatively common, and the vast majority of them are benign (Knox, 2013).
A thyroid nodule is a growth of cells (a lump) in the thyroid gland, which is located in the anterior neck region. In areas of the world where iodine is abundant, epidemiologic studies have revealed that the prevalence of palpable thyroid nodules is around 5% in women and 1% in males (Vander et al. Consequently, thyroid cancer affects women more commonly than men (3:1). Although thyroid cancer can affect people of any age, it is more common 1n individuals between the ages of 45 and 54, with a mean age of 50 at diagnosis (National Cancer Institute, 2015).