Browse Prior Art Database

MINING OF MULTIPLE PROTEIN ASSOCIATIONS FROM HISTOLOGICAL AND CELLULAR IMAGES

IP.com Disclosure Number: IPCOM000194149D
Publication Date: 2010-Mar-19
Document File: 9 page(s) / 40K

Publishing Venue

The IP.com Prior Art Database

Abstract

A technique to predict protein association which quantifies protein co-localization and co-expression is disclosed. This technique proposes measurement of co-localization of multi-proteins systems with statistical significance both at pixel and compartment level. The method also estimates co-expression of proteins that are located within different compartments. The proposed method provides a metric that can evaluate the co-localization and co-expression of different proteins at a compartment level. Confidence (MPA) scores, obtained from an algorithm, provide quantitative insight into the stored functional information and, thus can act as an indicator of clinical outcomes.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 25% of the total text.

Page 1 of 9

RP13318

MINING OF MULTIPLE PROTEIN ASSOCIATIONS FROM HISTOLOGICAL AND CELLULAR IMAGES

BRIEF ABSTRACT

    A technique to predict protein association which quantifies protein co- localization and co-expression is disclosed. This technique proposes measurement of co-localization of multi-proteins systems with statistical significance both at pixel and compartment level. The method also estimates co- expression of proteins that are located within different compartments. The proposed method provides a metric that can evaluate the co-localization and co- expression of different proteins at a compartment level. Confidence (MPA) scores, obtained from an algorithm, provide quantitative insight into the stored functional information and, thus can act as an indicator of clinical outcomes.

KEYWORDS

    Mining, multiple protein associations, histological, co-localization, co- expression, cellular compartment, quantitative, algorithm, co-localization coefficients, pixel, motifs, positive, negative, confidence scores

DETAILED DESCRIPTION

    Protein interactions are fundamental to understand both the functions of proteins and entire biological processes. Manual curation which utilizes more time and cost cannot keep up with the rapidly growing amount of literature and

1

Page 2 of 9

RP13318

the increasing number of newly discovered proteins. There is a need for text- mining tools to facilitate the extraction of such information.

    Generally, pair wise protein interactions which are manifested through their co-localization (expression of proteins within same cellular compartment) are studied through several methods. These methods automatically analyze expression of proteins from images. Co-localization of multiple proteins has been further analyzed at a pixel level. However, the conventional techniques fail to estimate the positive and negative association between multiple proteins in a multi-molecule framework.

    Conventional methods, quantify co-localization of two protein systems based on correlations or co-localization coefficients. To analyze multiple proteins co-localization systems each pair of proteins are considered individually or motifs are defined for assessing co-localization. The co-localization is often quantified by computing the Pearson's correlation coefficient. The correlation coefficient is a robust estimator for co-localization, but unfortunately it lacks biological relevance. Correlation measure was replaced by a more biological meaningful M-value or co-localization coefficient. The M-value quantifies the co-localized fraction of each molecular species, but it requires a threshold value for each channel. This value is then used as a cutoff between specific staining versus nonspecific staining. The overlapping regions between both channels that are above cutoff are then considered as co-localized regions, and the proportions of signal for each channel inside those areas are defined as co-localization coefficients. However, a problem with this...