피부섬유모세포 전사체 정보를 활용한구간 선택 기반 연령 예측 = Age Prediction based on the Transcriptome of Human Dermal Fibroblasts through Interval Selection

다국어 초록 (Multilingual Abstract)

It is reported that genome-wide RNA-seq profiles has potential as biomarkers of aging. A number of researchesachieved promising prediction performance based on gene expression profiles. We develop an age predictionmethod based on the transcriptome of human dermal fibroblasts by selecting a proper age interval. The proposedmethod executes multiple rules in a sequential manner and a rule utilizes a classifier and a regression model todetermine whether a given test sample belongs to the target age interval of the rule. If a given test sample satisfiesthe selection condition of a rule, age is predicted from the associated target age interval. Our method predicts ageto a mean absolute error of 5.7 years. Our method outperforms prior best performance of mean absolute error of7.7 years achieved by an ensemble based prediction method. We observe that it is possible to predict age basedon genome-wide RNA-seq profiles but prediction performance is not stable but varying with age.

국문 초록 (Abstract)

본 논문에서는 인간의 피부섬유모세포(Human dermal fibroblasts)로부터 확보한 전사체 정보를 활용하여 나이를 예측하는 방법을 소개한다. 제안 방법에서는 훈련을 통해 확보한 분류기 및 회귀 모델을 이용하여 샘플이 속한 적합한 연령 그룹을 선택한 후,선택된 연령 그룹에 속하는 훈련 데이터의 관측값을 활용하여 구체적인 연령을 예측한다. 연령을 예측하려는 샘플이 입력되면 복수개의 판별 규칙이 순서대로 실행되는데, 개별 판별 규칙에서는 분류기와 회귀 모델을 동시에 실행하여 해당 판별 규칙에 대한 선택조건이 만족되는지 여부를 확인한다. 선택 조건이 만족될 경우 판별 규칙의 타겟 연령 그룹에 속하는 데이터를 이용하여 훈련된 회귀 모델로 연령을 예측하며, 선택 조건이 만족되지 않으면 후속 판별 규칙을 실행한다. 공개 데이터에 대하여 실험한 결과 기존 연구에서 달성한 7.7년의 평균 예측 오차보다 우수한 5.7년이라는 평균 예측 오차를 달성함을 확인하였다.

참고문헌 (Reference)

1 Julia Tigges, "The hallmarks of fibroblast ageing" Elsevier BV 138 : 26-44, 2014

2 Aaron M. Smith, "Standard machine learning approaches outperform deep representation learning on phenotype prediction from transcriptomics data" Springer Science and Business Media LLC 21 (21): 119-, 2020

3 F. Pedregosa, "Scikit-Learn : Machine Learning in Python" 12 : 2825-2830, 2011

4 A. Drouin, "Predictive Computational Phenotyping and Biomarker Discovery using Reference-free Genome Comparison" 17 : 754-, 2016

5 Jason G. Fleischer, "Predicting age from the transcriptome of human dermal fibroblasts" Springer Science and Business Media LLC 19 (19): 221-, 2018

6 K. P. Murphy, "Machine learning: a probabilistic perspective" The MIT Press 492-493, 2012

7 A. Drouin, "Interpretable Genotype-tophenotype Classifiers with Performance Guarantees" 9 : 4071-, 2019

8 Philip N. Benfey, "From Genotype to Phenotype: Systems Biology Meets Natural Variation" American Association for the Advancement of Science (AAAS) 320 (320): 495-497, 2008

9 Renata Zbieć-Piekarska, "Development of a forensically useful age prediction method based on DNA methylation analysis" Elsevier BV 17 : 173-179, 2015

10 Zhenyu Tang, "Deep Learning of Imaging Phenotype and Genotype for Predicting Overall Survival Time of Glioblastoma Patients" Institute of Electrical and Electronics Engineers (IEEE) 39 (39): 2100-2109, 2020

11 Alexander I. Young, "Deconstructing the sources of genotype-phenotype associations in humans" American Association for the Advancement of Science (AAAS) 365 (365): 1396-1400, 2019

12 Steve Horvath, "DNA methylation age of human tissues and cell types" Springer Science and Business Media LLC 14 (14): 3156-, 2013

13 J. Phillip, "Biophysical and Biomolecular Determination of Cellular Age in Humans" 1 : 2017

14 Yoav Freund, "A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting" Elsevier BV 55 (55): 119-139, 1997