Classification of protein profiles using fuzzy clustering techniques: an application in early diagnosis of oral, cervical and ovarian cancer

TY - GEN

T1 - Classification of protein profiles using fuzzy clustering techniques

T2 - 32nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society

AU - Karemore, Gopal

AU - Mullick, Jhinuk B.

AU - Sujatha, R.

AU - Nielsen, Mads

AU - Santhosh, C.

N1 - Conference code: 32

PY - 2010

Y1 - 2010

N2 - Present study has brought out a comparison of PCA and fuzzy clustering techniques in classifying protein profiles (chromatogram) of homogenates of different tissue origins: Ovarian, Cervix, Oral cancers, which were acquired using HPLC-LIF (High Performance Liquid Chromatography-Laser Induced Fluorescence) method developed in our laboratory. Study includes 11 chromatogram spectra each from oral, cervical, ovarian cancers as well as healthy volunteers. Generally multivariate analysis like PCA demands clear data that is devoid of day-to-day variation, artifacts due to experimental strategies, inherent uncertainty in pumping procedure which is very common activities during HPLC-LIF experiment. Under these circumstances we demonstrate how fuzzy clustering algorithm like Gath Geva followed by Sammon mapping outperform PCA mapping in classifying various cancers from healthy spectra with classification rate up to 95 % from 60%. Methods are validated using various clustering indexes and shows promising improvement in developing optical pathology like HPLC-LIF for early detection of various cancers in all uncertain conditions with high sensitivity and specificity.

AB - Present study has brought out a comparison of PCA and fuzzy clustering techniques in classifying protein profiles (chromatogram) of homogenates of different tissue origins: Ovarian, Cervix, Oral cancers, which were acquired using HPLC-LIF (High Performance Liquid Chromatography-Laser Induced Fluorescence) method developed in our laboratory. Study includes 11 chromatogram spectra each from oral, cervical, ovarian cancers as well as healthy volunteers. Generally multivariate analysis like PCA demands clear data that is devoid of day-to-day variation, artifacts due to experimental strategies, inherent uncertainty in pumping procedure which is very common activities during HPLC-LIF experiment. Under these circumstances we demonstrate how fuzzy clustering algorithm like Gath Geva followed by Sammon mapping outperform PCA mapping in classifying various cancers from healthy spectra with classification rate up to 95 % from 60%. Methods are validated using various clustering indexes and shows promising improvement in developing optical pathology like HPLC-LIF for early detection of various cancers in all uncertain conditions with high sensitivity and specificity.

KW - Former Faculty of Life Sciences

U2 - 10.1109/IEMBS.2010.5627292

DO - 10.1109/IEMBS.2010.5627292

M3 - Article in proceedings

SN - 978-1-4244-4123-5

T3 - I E E E Engineering in Medicine and Biology Society. Conference Proceedings

SP - 6361

EP - 6364

BT - 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)

PB - IEEE

Y2 - 31 August 2010 through 4 September 2010

ER -