마이크로어레이 데이터와 PPI 데이터를 이용한 에스트로겐 수용체 음성 유방암 환자의 예후 특이 네트워크 식별 및 예후 예측

황유현(Youhyeon Hwang),오민(Min Oh),윤영미(Youngmi Yoon) 한국컴퓨터정보학회 2015 韓國컴퓨터情報學會論文誌 Vol.20 No.2

본 논문에서는 유전자 네트워크를 기반으로 유방암 환자의 예후를 예측하는 알고리듬을 제안한다. 유방암 환자의 마이크로어레이 데이터와 PPI(Protein-protein interaction)데이터를 이용하여 알고리듬의 분류자로 사용될 예후 특이 네트워크(Prognosis specific gene network)를 추출한다. PPI에 속한 모든 유전자 네트워크에 대하여 각각의 네트워크가 예후 좋음과 나쁨을 잘 구분하는지에 대한 점수를 피어슨 상관계수(Pearson"s correlation coefficient)와 마이크로어레이 데이터를 이용하여 계산한다. 이들 중 가장 예후에 유의한 네트워크를 식별하고, 이 네트워크를 분류자로 사용하여 에스트로겐 수용체 음성 유방암 환자의 예후를 분류 분석 한다. 본 연구와 기존 연구의 알고리듬 정확도를 비교 분석 하기 위하여 독립 실험을 진행하고, 본 연구에서 제안된 알고리듬의 성능이 더 우수함을 보인다. 또한, Gene Ontology 데이터베이스를 활용하여 식별된 예후 특이 네트워크를 기능적으로 검증 한다. This study proposes an algorithm for predicting breast cancer prognosis based on genetic network. We identify prognosis-specific network using gene expression data and PPI(protein-protein interaction) data. To acquire the network, we calculate Pearson"s correlation coefficient(PCC) between genes in all PPI pairs using gene expression data. We develop a prediction model for breast cancer patients with estrogen-receptor-negative using the network as a classifier. We compare classification performance of our algorithm with existing algorithms on independent data and shows our algorithm is improved. In addition, we make an functionality analysis on the genes in the prognosis-specific network using GO(Gene Ontology) enrichment validation.

Extraction of specific common genetic network of side effect pair, and prediction of side effects for a drug based on PPI network

Youhyeon Hwang(황유현),Min Oh(오민),Youngmi Yoon(윤영미) 한국컴퓨터정보학회 2016 韓國컴퓨터情報學會論文誌 Vol.21 No.1

In this study, we collect various side effect pairs which are appeared frequently at many drugs, and select side effect pairs that have higher severity. For every selected side effect pair, we extract common genetic networks which are shared by side effects’ genes and drugs’ target genes based on PPI(Protein-Protein Interaction) network. For this work, firstly, we gather drug related data, side effect data and PPI data. Secondly, for extracting common genetic network, we find shortest paths between drug target genes and side effect genes based on PPI network, and integrate these shortest paths. Thirdly, we develop a classification model which uses this common genetic network as a classifier. We calculate similarity score between the common genetic network and genetic network of a drug for classifying the drug. Lastly, we validate our classification model by means of AUC(Area Under the Curve) value.

약물-질병 경로 네트워크의 구축과 비교를 통한 신약재창출

황소윤(Soyoun Hwang),황유현(Youhyeon Hwang),오민(Min Oh),윤영미(Youngmi Yoon) 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.9

As genome data are continuously increasing, there have been various efforts that elucidate a mechanism of drug action. However, the mechanism of drug action is still not elucidated perfectly despite various previous efforts. In this paper, we constructed drug-disease gene network based on protein-protein interaction to better understand a mechanism of drug action, and applied drug repositioning. In order to construct pathway networks for drug-disease pairs, we connected the drug target genes, drug related genes and the disease related genes respectively. We extracted critical gene networks for particular drug-disease pairs of which therapeutic indications are known. By estimating similarity between the critical gene networks and candidate networks derived from drug-disease pairs whose therapeutic effects are unknown, new drug indications were predicted. The putative drug indications significantly overlapped with those of CTD database (Fishers exact P = 7.01E-104). In performance evaluation, the prediction result showed the area under the ROC curve (AUC) of 0.7122.

Target Prediction Based On PPI Network

Taekeon Lee(이태건),Youhyeon Hwang(황유현),Min Oh(오민),Youngmi Yoon(윤영미) 韓國컴퓨터情報學會 2016 韓國컴퓨터情報學會論文誌 Vol.21 No.3

To reduce the expenses for development a novel drug, systems biology has been studied actively. Target prediction, a part of systems biology, contributes to finding a new purpose for FDA(Food and Drug Administration) approved drugs and development novel drugs. In this paper, we propose a classification model for predicting novel target genes based on relation between target genes and disease related genes. After collecting known target genes from TTD(Therapeutic Target Database) and disease related genes from OMIM(Online Mendelian Inheritance in Man), we analyzed the effect of target genes on disease related genes based on PPI(Protein-Protein Interactions) network. We focused on the distinguishing characteristics between known target genes and random target genes, and used the characteristics as features for building a classifier. Because our model is constructed using information about only a disease and its known targets, the model can be applied to unusual diseases without similar drugs and diseases, while existing models for finding new drug-disease associations are based on drug-drug similarity and disease-disease similarity. We validated accuracy of the model using LOOCV of ten times and the AUCs were 0.74 on Alzheimer’s disease and 0.71 on Breast cancer.

유사 약물 예측을 위한 텍스트 마이닝 기반의 새로운 약물 유사도 측정 방법

장기업(Giup Jang),황유현(Youhyeon Hwang),오민(Min Oh),이태건(Taekeon Lee),윤영미(Youngmi Yoon) 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.7

There are many studies to identify new indications of existing drugs. Among them, text mining derives new relationships between drug and disease from vast unstructured data. In this study, co-occurrence of drug and disease in a sentences of abstracts in literatures is counted, and also co-occurrence of drug and gene is counted. Then drug-disease co-occurrence matrix and drug-gene co-occurrence matrix are generated. For each drug pairs using these matrices, disease-based drug similarity and gene-based drug similarity are calculated by mutual information. Also chemical, side-effect, and GO similarity are calculated for each drug pairs, respectively. Classification class label of “same” is given to each drug pair if ATC code for the two drugs of the pair is equal, and “different” is given to the pair otherwise. For the classification of drug pairs, AUC is improved with addition of mutual information, and we validated that text mining can help identifying similar drug. Identification of similar drugs can be utilized for drug repositioning.

통합 마이크로어레이 데이터를 이용한 유방암의 분자적 서브타입의 분류

이지연(Jeeyeon Lee),오민(Min Oh),황유현(Youhyeon Hwang),윤영미(Youngmi Yoon) 한국정보기술학회 2015 한국정보기술학회논문지 Vol.13 No.2

There are molecular sub-types in breast cancer which are Basal-like, Luminal A, Luminal B, and ERBB2/HER2+. These subtypes are major factors to prognosis. Prediction of correct sub-type helps doctors to choose best treatment for patient, and is critical factor which affects prognosis of patient consequently. However, predicting the class of sub-types is not easy due to multi-class classification, accordingly the classifiers of existing studies show low accuracy. Microarray data contains many gene expression, and has been used for predicting phenotypes of numerous cancers. However, the number of samples in each microarray data set is limited. Increasing the number of data by integrating data is important for high classification accuracy. In our study, we integrate data by ranking the gene expression values, and build sub-type classifiers using NaiveBayes, Lib-SVM, C4.5. We run gene ontology enrichment analysis with small number of informative genes which still give high classification accuracy. Among 65 Go-Terms which are statistically significant (p-value < 0.01), we identified 11 cancer relating GO-terms, and suggested a set of genes which belong to those GO-terms.

통합 마이크로어레이 데이터를 이용한 유방암-서브타입의 분류

이지연(JeeYeon Lee),오민(Min Oh),황유현(YouHyeon Hwang),윤영미(Youngmi Yoon) 한국정보기술학회 2014 Proceedings of KIIT Conference Vol.2014 No.5

유방암의 분자적 서브타입(Molecular Sub-type)은 대표적으로 Basal-like Luminal A, Luminal B, ERBB2/HER2+이 있다. 환자의 분자적 서브타입을 정확하게 진단하는 것은 임상에서 최적의 치료방법을 선택할 수 있도록 하며 결과적으로 환자의 예후에 많은 영향을 미치는 매우 중요한 요소이다. 그러나 생물학적 이질성으로 인해 다양한 유방암의 분자적 서브타입을 예측하기에는 큰 어려움이 있으며 기존 연구에서 제안되었던 분류자들 또한 높은 정확도를 구하기 쉽지 않았다. 마이크로어레이 데이터는 적게는 수천 개 많게는 수만 개의 유전자개의 발현 값을 포함하고 있어 다양한 암의 질병의 발현 형질 분류에 유용하게 쓰인다 하지만 독립으로 얻어진 마이크로어레이 데이터의 샘플 수가 제한적이기 때문에 정확한 분석을 위해 여러 개의 마이크로어레이 데이터를 통합하여 샘플 수를 늘리는 것은 중요한 요소이다. 본 논문에서는 유전자 발현량 순위를 이용하여 이질적인 마이크로어레이들을 효율적으로 정규화하고 통합한다 그리고 통합 데이터와 Naive Bayes 분류기를 사용해 보다 정확히 분자 유형을 분류한다. There are 4 typical sub-types in breast cancer which are Basal-like, Luminal A, Luminal B,and ERBB2/HER2+. Diagnosis of correct sub-type helps doctors to choose best treatment for each patient, and is critical factor which affects prognosis of patient consequently. However, predicting the class of molecular sub-types of breast cancer is not easy due to huge biological differences in each sub-type, and accordingly the classifiers of existing studies show low accuracy. Microarray data contains expression values of more than 10,000 genes, and has been used for predicting phenotypes of numerous cancers. However, the number of samples in each microarray data set is limited. Increasing the number of samples by integrating data sets is important for high classification accuracy. In our study, we integrate independent microarray data sets by ranking the expression values, and build a sub-type classifier using NaiveBayes.