Comparison of Methods for Feature Selection in Clustering of High-Dimensional RNA-Sequencing Data to Identify Cancer Subtypes.
Comparison of Methods for Feature Selection in Clustering of High-Dimensional RNA-Sequencing Data to Identify Cancer Subtypes.
复制标题
高维RNA测序数据聚类中特征选择方法的比较，以识别癌症亚型。
DOI：
10.3389/fgene.2021.632620
复制
发表时间：
2021
影响因子：
3.7
通讯作者：
Rydén P
中科院分区：
文献类型：
--
作者：
Källberg D;Vidman L;Rydén P
癌症亚型鉴定对于促进癌症诊断和选择有效治疗是重要的。基于高维RNA测序数据的癌症患者聚类可用于检测新亚型，但仅检测特征的子集（例如，基因）包含与癌症亚型相关的信息。因此，可以合理地假设聚类应该基于一组精心选择的特征，而不是所有特征。已经提出了几种特征选择方法，但如何以及何时使用这些方法仍然知之甚少。在四个人类癌症数据集上评估了十三种特征选择方法，所有这些数据集都具有已知的亚型（金标准），这些亚型仅用于评估。通过考虑所选基因的平均表达和标准差（SD）、与其他方法的重叠以及它们的聚类性能来表征所述方法，使用调整的兰德指数（ARI）将聚类结果与金标准进行比较。将结果与作为阳性对照的监督方法和两个阴性对照进行比较，其中包括随机选择的基因或所有基因。对于所有数据集，最佳特征选择方法优于阴性对照，对于两个数据集，增益显著，ARI分别从（-0.01，0.39）增加到（0.66，0.72）。没有一种特征选择方法完全优于其他方法，但使用dip-rest统计量选择1000个基因总体上是一个不错的选择。通常使用的方法，即选择具有最高SD的基因，在我们的研究中表现不佳。
Cancer subtype identification is important to facilitate cancer diagnosis and select effective treatments. Clustering of cancer patients based on high-dimensional RNA-sequencing data can be used to detect novel subtypes, but only a subset of the features (e.g., genes) contains information related to the cancer subtype. Therefore, it is reasonable to assume that the clustering should be based on a set of carefully selected features rather than all features. Several feature selection methods have been proposed, but how and when to use these methods are still poorly understood. Thirteen feature selection methods were evaluated on four human cancer data sets, all with known subtypes (gold standards), which were only used for evaluation. The methods were characterized by considering mean expression and standard deviation (SD) of the selected genes, the overlap with other methods and their clustering performance, obtained comparing the clustering result with the gold standard using the adjusted Rand index (ARI). The results were compared to a supervised approach as a positive control and two negative controls in which either a random selection of genes or all genes were included. For all data sets, the best feature selection approach outperformed the negative control and for two data sets the gain was substantial with ARI increasing from (−0.01, 0.39) to (0.66, 0.72), respectively. No feature selection method completely outperformed the others but using the dip-rest statistic to select 1000 genes was overall a good choice. The commonly used approach, where genes with the highest SDs are selected, did not perform well in our study.
登录
查看更多内容
影响因子：
3
作者：
Freyhult E;Landfors M;Önskog J;Hvidsten TR;Rydén P
通讯作者：
Rydén P
DOI：
10.1056/nejmoa1402121
发表时间：
2015-06-25
期刊：
The New England journal of medicine
影响因子：
--
作者：
Cancer Genome Atlas Research Network;Brat DJ;Verhaak RG;Aldape KD;Yung WK;Salama SR;Cooper LA;Rheinbay E;Miller CR;Vitucci M;Morozova O;Robertson AG;Noushmehr H;Laird PW;Cherniack AD;Akbani R;Huse JT;Ciriello G;Poisson LM;Barnholtz-Sloan JS;Berger MS;Brennan C;Colen RR;Colman H;Flanders AE;Giannini C;Grifford M;Iavarone A;Jain R;Joseph I;Kim J;Kasaian K;Mikkelsen T;Murray BA;O'Neill BP;Pachter L;Parsons DW;Sougnez C;Sulman EP;Vandenberg SR;Van Meir EG;von Deimling A;Zhang H;Crain D;Lau K;Mallery D;Morris S;Paulauskis J;Penny R;Shelton T;Sherman M;Yena P;Black A;Bowen J;Dicostanzo K;Gastier-Foster J;Leraas KM;Lichtenberg TM;Pierson CR;Ramirez NC;Taylor C;Weaver S;Wise L;Zmuda E;Davidsen T;Demchok JA;Eley G;Ferguson ML;Hutter CM;Mills Shaw KR;Ozenberger BA;Sheth M;Sofia HJ;Tarnuzzer R;Wang Z;Yang L;Zenklusen JC;Ayala B;Baboud J;Chudamani S;Jensen MA;Liu J;Pihl T;Raman R;Wan Y;Wu Y;Ally A;Auman JT;Balasundaram M;Balu S;Baylin SB;Beroukhim R;Bootwalla MS;Bowlby R;Bristow CA;Brooks D;Butterfield Y;Carlsen R;Carter S;Chin L;Chu A;Chuah E;Cibulskis K;Clarke A;Coetzee SG;Dhalla N;Fennell T;Fisher S;Gabriel S;Getz G;Gibbs R;Guin R;Hadjipanayis A;Hayes DN;Hinoue T;Hoadley K;Holt RA;Hoyle AP;Jefferys SR;Jones S;Jones CD;Kucherlapati R;Lai PH;Lander E;Lee S;Lichtenstein L;Ma Y;Maglinte DT;Mahadeshwar HS;Marra MA;Mayo M;Meng S;Meyerson ML;Mieczkowski PA;Moore RA;Mose LE;Mungall AJ;Pantazi A;Parfenov M;Park PJ;Parker JS;Perou CM;Protopopov A;Ren X;Roach J;Sabedot TS;Schein J;Schumacher SE;Seidman JG;Seth S;Shen H;Simons JV;Sipahimalani P;Soloway MG;Song X;Sun H;Tabak B;Tam A;Tan D;Tang J;Thiessen N;Triche T Jr;Van Den Berg DJ;Veluvolu U;Waring S;Weisenberger DJ;Wilkerson MD;Wong T;Wu J;Xi L;Xu AW;Yang L;Zack TI;Zhang J;Aksoy BA;Arachchi H;Benz C;Bernard B;Carlin D;Cho J;DiCara D;Frazer S;Fuller GN;Gao J;Gehlenborg N;Haussler D;Heiman DI;Iype L;Jacobsen A;Ju Z;Katzman S;Kim H;Knijnenburg T;Kreisberg RB;Lawrence MS;Lee W;Leinonen K;Lin P;Ling S;Liu W;Liu Y;Liu Y;Lu Y;Mills G;Ng S;Noble MS;Paull E;Rao A;Reynolds S;Saksena G;Sanborn Z;Sander C;Schultz N;Senbabaoglu Y;Shen R;Shmulevich I;Sinha R;Stuart J;Sumer SO;Sun Y;Tasman N;Taylor BS;Voet D;Weinhold N;Weinstein JN;Yang D;Yoshihara K;Zheng S;Zhang W;Zou L;Abel T;Sadeghi S;Cohen ML;Eschbacher J;Hattab EM;Raghunathan A;Schniederjan MJ;Aziz D;Barnett G;Barrett W;Bigner DD;Boice L;Brewer C;Calatozzolo C;Campos B;Carlotti CG Jr;Chan TA;Cuppini L;Curley E;Cuzzubbo S;Devine K;DiMeco F;Duell R;Elder JB;Fehrenbach A;Finocchiaro G;Friedman W;Fulop J;Gardner J;Hermes B;Herold-Mende C;Jungk C;Kendler A;Lehman NL;Lipp E;Liu O;Mandt R;McGraw M;Mclendon R;McPherson C;Neder L;Nguyen P;Noss A;Nunziata R;Ostrom QT;Palmer C;Perin A;Pollo B;Potapov A;Potapova O;Rathmell WK;Rotin D;Scarpace L;Schilero C;Senecal K;Shimmel K;Shurkhay V;Sifri S;Singh R;Sloan AE;Smolenski K;Staugaitis SM;Steele R;Thorne L;Tirapelli DP;Unterberg A;Vallurupalli M;Wang Y;Warnick R;Williams F;Wolinsky Y;Bell S;Rosenberg M;Stewart C;Huang F;Grimsby JL;Radenbaugh AJ;Zhang J
通讯作者：
Zhang J
影响因子：
4.9
作者：
Fujikado, Noriyuki;Saijo, Shinobu;Iwakura, Yoichiro
通讯作者：
Iwakura, Yoichiro
影响因子：
3.7
作者：
Bentink S;Haibe-Kains B;Risch T;Fan JB;Hirsch MS;Holton K;Rubio R;April C;Chen J;Wickham-Garcia E;Liu J;Culhane A;Drapkin R;Quackenbush J;Matulonis UA
通讯作者：
Matulonis UA
影响因子：
11.2
作者：
Bertucci, F;Finetti, P;Birnbaum, D
通讯作者：
Birnbaum, D