Analysis of mRNA expression profiles of 4332 drug targets across 68 major cancer types: A new bioinformatic method to predict dependency of tumors on drug targets

John Mpindi, Henri Sara, Sami Kilpinen, Tommi Pisto, Elmar Bucher, Kalle Ojala, Kristiina Iljin, Saija Haapa-Paananen, Heikki Joensuu, Matthias Nees, Olli Kallioniemi

Research output: Contribution to conferenceConference articleScientific

Abstract

Bringing a new drug from discovery to the clinic takes a decade, is very expensive and often fails due to efficacy or safety concerns. Many drugs are developed against specific molecular targets and mechanisms which may be shared across a wide variety of tumor types. Therefore, systematic identification of tumor types and subtypes, where these targets play a particularly important role would be highly informative. Here, we have developed a new bioinformatic strategy towards the identification of tumor dependencies on individual drug targets. We applied our annotated database of mRNA expression data from 9,783 samples across 43 normal human tissue types and 68 major cancer types as a basis for these studies (www.genesapiens.org). We modified a statistical method from economics and adapted this for gene expression meta-analyses and calculated Gene Tissue outlier Indices (GTI) for all genes. GTI scores highlight subsets of tumor types where expression of a drug target is higher than in any other normal and malignant tissue included in the database. GTI scores were analyzed for 4332 known and emerging drug targets. This resulted in the identification of 494 statistically significant associations between a particular tumor type and a drug target. Among the strongest outliers were known drug target-disease combinations, such as KIT in gastrointestinal stromal tumors and FLT3 in AML and ALL. This highlights the known potential of Gleevec in GIST and emerging potential of FLT3 inhibitor Sunitinib in leukemias. In summary, we have developed a new bioinformatic method to predict the dependency of tumors on a given drug target. This is based on a statistical method, where expression levels of mRNA are compared not just between tumors and normal tissues in a specific organ, but also across all normal and tumor tissues. The GTI scores indicated many potential new therapeutic opportunities to explore in preclinical and clinical studies.
Original languageEnglish
Publication statusPublished - 2009
MoE publication typeNot Eligible
Event100th Annual Meeting of the American Association for Cancer Research - Denver, United States
Duration: 18 Apr 200922 Apr 2009

Conference

Conference100th Annual Meeting of the American Association for Cancer Research
CountryUnited States
CityDenver
Period18/04/0922/04/09

Fingerprint

Computational Biology
Messenger RNA
Pharmaceutical Preparations
Neoplasms
Genes
Databases
Gastrointestinal Stromal Tumors
Drug Discovery
Meta-Analysis
Leukemia
Economics
Safety
Gene Expression

Cite this

Mpindi, J., Sara, H., Kilpinen, S., Pisto, T., Bucher, E., Ojala, K., ... Kallioniemi, O. (2009). Analysis of mRNA expression profiles of 4332 drug targets across 68 major cancer types: A new bioinformatic method to predict dependency of tumors on drug targets. Paper presented at 100th Annual Meeting of the American Association for Cancer Research, Denver, United States.
Mpindi, John ; Sara, Henri ; Kilpinen, Sami ; Pisto, Tommi ; Bucher, Elmar ; Ojala, Kalle ; Iljin, Kristiina ; Haapa-Paananen, Saija ; Joensuu, Heikki ; Nees, Matthias ; Kallioniemi, Olli. / Analysis of mRNA expression profiles of 4332 drug targets across 68 major cancer types : A new bioinformatic method to predict dependency of tumors on drug targets. Paper presented at 100th Annual Meeting of the American Association for Cancer Research, Denver, United States.
@conference{8608726f90884c968e6d4a7f0d611f64,
title = "Analysis of mRNA expression profiles of 4332 drug targets across 68 major cancer types: A new bioinformatic method to predict dependency of tumors on drug targets",
abstract = "Bringing a new drug from discovery to the clinic takes a decade, is very expensive and often fails due to efficacy or safety concerns. Many drugs are developed against specific molecular targets and mechanisms which may be shared across a wide variety of tumor types. Therefore, systematic identification of tumor types and subtypes, where these targets play a particularly important role would be highly informative. Here, we have developed a new bioinformatic strategy towards the identification of tumor dependencies on individual drug targets. We applied our annotated database of mRNA expression data from 9,783 samples across 43 normal human tissue types and 68 major cancer types as a basis for these studies (www.genesapiens.org). We modified a statistical method from economics and adapted this for gene expression meta-analyses and calculated Gene Tissue outlier Indices (GTI) for all genes. GTI scores highlight subsets of tumor types where expression of a drug target is higher than in any other normal and malignant tissue included in the database. GTI scores were analyzed for 4332 known and emerging drug targets. This resulted in the identification of 494 statistically significant associations between a particular tumor type and a drug target. Among the strongest outliers were known drug target-disease combinations, such as KIT in gastrointestinal stromal tumors and FLT3 in AML and ALL. This highlights the known potential of Gleevec in GIST and emerging potential of FLT3 inhibitor Sunitinib in leukemias. In summary, we have developed a new bioinformatic method to predict the dependency of tumors on a given drug target. This is based on a statistical method, where expression levels of mRNA are compared not just between tumors and normal tissues in a specific organ, but also across all normal and tumor tissues. The GTI scores indicated many potential new therapeutic opportunities to explore in preclinical and clinical studies.",
author = "John Mpindi and Henri Sara and Sami Kilpinen and Tommi Pisto and Elmar Bucher and Kalle Ojala and Kristiina Iljin and Saija Haapa-Paananen and Heikki Joensuu and Matthias Nees and Olli Kallioniemi",
year = "2009",
language = "English",
note = "100th Annual Meeting of the American Association for Cancer Research ; Conference date: 18-04-2009 Through 22-04-2009",

}

Mpindi, J, Sara, H, Kilpinen, S, Pisto, T, Bucher, E, Ojala, K, Iljin, K, Haapa-Paananen, S, Joensuu, H, Nees, M & Kallioniemi, O 2009, 'Analysis of mRNA expression profiles of 4332 drug targets across 68 major cancer types: A new bioinformatic method to predict dependency of tumors on drug targets' Paper presented at 100th Annual Meeting of the American Association for Cancer Research, Denver, United States, 18/04/09 - 22/04/09, .

Analysis of mRNA expression profiles of 4332 drug targets across 68 major cancer types : A new bioinformatic method to predict dependency of tumors on drug targets. / Mpindi, John; Sara, Henri; Kilpinen, Sami; Pisto, Tommi; Bucher, Elmar; Ojala, Kalle; Iljin, Kristiina; Haapa-Paananen, Saija; Joensuu, Heikki; Nees, Matthias; Kallioniemi, Olli.

2009. Paper presented at 100th Annual Meeting of the American Association for Cancer Research, Denver, United States.

Research output: Contribution to conferenceConference articleScientific

TY - CONF

T1 - Analysis of mRNA expression profiles of 4332 drug targets across 68 major cancer types

T2 - A new bioinformatic method to predict dependency of tumors on drug targets

AU - Mpindi, John

AU - Sara, Henri

AU - Kilpinen, Sami

AU - Pisto, Tommi

AU - Bucher, Elmar

AU - Ojala, Kalle

AU - Iljin, Kristiina

AU - Haapa-Paananen, Saija

AU - Joensuu, Heikki

AU - Nees, Matthias

AU - Kallioniemi, Olli

PY - 2009

Y1 - 2009

N2 - Bringing a new drug from discovery to the clinic takes a decade, is very expensive and often fails due to efficacy or safety concerns. Many drugs are developed against specific molecular targets and mechanisms which may be shared across a wide variety of tumor types. Therefore, systematic identification of tumor types and subtypes, where these targets play a particularly important role would be highly informative. Here, we have developed a new bioinformatic strategy towards the identification of tumor dependencies on individual drug targets. We applied our annotated database of mRNA expression data from 9,783 samples across 43 normal human tissue types and 68 major cancer types as a basis for these studies (www.genesapiens.org). We modified a statistical method from economics and adapted this for gene expression meta-analyses and calculated Gene Tissue outlier Indices (GTI) for all genes. GTI scores highlight subsets of tumor types where expression of a drug target is higher than in any other normal and malignant tissue included in the database. GTI scores were analyzed for 4332 known and emerging drug targets. This resulted in the identification of 494 statistically significant associations between a particular tumor type and a drug target. Among the strongest outliers were known drug target-disease combinations, such as KIT in gastrointestinal stromal tumors and FLT3 in AML and ALL. This highlights the known potential of Gleevec in GIST and emerging potential of FLT3 inhibitor Sunitinib in leukemias. In summary, we have developed a new bioinformatic method to predict the dependency of tumors on a given drug target. This is based on a statistical method, where expression levels of mRNA are compared not just between tumors and normal tissues in a specific organ, but also across all normal and tumor tissues. The GTI scores indicated many potential new therapeutic opportunities to explore in preclinical and clinical studies.

AB - Bringing a new drug from discovery to the clinic takes a decade, is very expensive and often fails due to efficacy or safety concerns. Many drugs are developed against specific molecular targets and mechanisms which may be shared across a wide variety of tumor types. Therefore, systematic identification of tumor types and subtypes, where these targets play a particularly important role would be highly informative. Here, we have developed a new bioinformatic strategy towards the identification of tumor dependencies on individual drug targets. We applied our annotated database of mRNA expression data from 9,783 samples across 43 normal human tissue types and 68 major cancer types as a basis for these studies (www.genesapiens.org). We modified a statistical method from economics and adapted this for gene expression meta-analyses and calculated Gene Tissue outlier Indices (GTI) for all genes. GTI scores highlight subsets of tumor types where expression of a drug target is higher than in any other normal and malignant tissue included in the database. GTI scores were analyzed for 4332 known and emerging drug targets. This resulted in the identification of 494 statistically significant associations between a particular tumor type and a drug target. Among the strongest outliers were known drug target-disease combinations, such as KIT in gastrointestinal stromal tumors and FLT3 in AML and ALL. This highlights the known potential of Gleevec in GIST and emerging potential of FLT3 inhibitor Sunitinib in leukemias. In summary, we have developed a new bioinformatic method to predict the dependency of tumors on a given drug target. This is based on a statistical method, where expression levels of mRNA are compared not just between tumors and normal tissues in a specific organ, but also across all normal and tumor tissues. The GTI scores indicated many potential new therapeutic opportunities to explore in preclinical and clinical studies.

M3 - Conference article

ER -

Mpindi J, Sara H, Kilpinen S, Pisto T, Bucher E, Ojala K et al. Analysis of mRNA expression profiles of 4332 drug targets across 68 major cancer types: A new bioinformatic method to predict dependency of tumors on drug targets. 2009. Paper presented at 100th Annual Meeting of the American Association for Cancer Research, Denver, United States.