Home > Riviste > Minerva Biotecnologica > Fascicoli precedenti > Minerva Biotecnologica 2002 December;14(3-4) > Minerva Biotecnologica 2002 December;14(3-4):281-90

ULTIMO FASCICOLO
 

ARTICLE TOOLS

Estratti

MINERVA BIOTECNOLOGICA

Rivista di Biologia Molecolare e Biotecnologie


Indexed/Abstracted in: EMBASE, Science Citation Index Expanded (SciSearch), Scopus
Impact Factor 0,25


eTOC

 

  MICROARRAY MEETING 2002: NEW DEVELOPMENTS IN MUTATION DETECTION AND GENE EXPRESSION
Segrate, MI (Italy), April 12, 2002
FREEfree


Minerva Biotecnologica 2002 December;14(3-4):281-90

Copyright © 2003 EDIZIONI MINERVA MEDICA

lingua: Inglese

Disjoint PCA models for marker identification and classification of cancer types using gene expression data

Bicciato S., Luchini A., Di Bello C.

Department of Chemical Process Engineering, University of Padova, Padova


FULL TEXT  


Back­ground. The par­allel mon­i­tor­ing of the expres­sion pro­files of thou­sands of ­genes ­seems par­tic­u­lar­ly prom­is­ing for a deep­er under­stand­ing of can­cer biol­o­gy and to iden­ti­fy molec­u­lar sig­na­tures sup­port­ing the his­to­log­i­cal clas­sifi­ca­tion ­schemes of neo­plas­tic spec­i­mens. How­ev­er, molec­u­lar diag­nos­tic ­based on micro­ar­ray ­data ­presents ­major chal­leng­es due to the com­plex, mul­ti­class ­nature and to the over­whelm­ing num­ber of var­i­ables char­ac­ter­iz­ing ­gene expres­sion data­bas­es of mul­ti­ple ­tumor sam­ples. ­Thus, the devel­op­ment of mul­ti­class clas­sifi­ca­tion ­schemes and of mark­er selec­tion meth­ods, ­that ­allow the simul­ta­ne­ous clas­sifi­ca­tion of mul­ti­ple ­tumor ­types and the iden­tifi­ca­tion of ­those ­genes ­that are ­most like­ly to con­fer ­high clas­sifi­ca­tion accu­ra­cy, is of par­a­mount impor­tance.
Meth­ods. A com­pu­ta­tion­al pro­ce­dure for mark­er iden­tifi­ca­tion and clas­sifi­ca­tion of mul­ti­class ­gene expres­sion ­data ­through the appli­ca­tion of dis­joint prin­ci­pal com­po­nent mod­els, ­based on the ­Soft Inde­pen­dent Mod­el­ing of ­Class Anal­o­gy ­approach (SIM­CA), is ­described. The iden­ti­fied fea­tures rep­re­sent a ration­al and dimen­sion­al­ly ­reduced ­base for under­stand­ing the ­basic biol­o­gy of dis­eas­es, defin­ing tar­gets of ther­a­peu­tic inter­ven­tion, and devel­op­ing diag­nos­tic ­tools for the iden­tifi­ca­tion and clas­sifi­ca­tion of mul­ti­ple path­o­log­i­cal ­states.
­Results. The meth­od has ­been test­ed on 2 dif­fer­ent micro­ar­ray ­data ­sets ­obtained ­from var­i­ous ­human ­tumor sam­ples: i) ­acute leu­ke­mi­as, and ii) ­small ­round ­blue-­cell ­tumors.
Con­clu­sions. The ­results dem­on­strate ­that the dis­joint PCA mod­el­ing pro­ce­dure ­allows the iden­tifi­ca­tion of spe­cif­ic phe­no­type mark­ers and pro­vides the assign­ment to mul­ti­ple class­es for pre­vi­ous­ly ­unseen instanc­es.

inizio pagina

Publication History

Per citare questo articolo

Corresponding author e-mail