Multidimensional algorithm: Cancer detector

Skip to Navigation


  • Published: Aug 15, 2016
  • Author: David Bradley
  • Channels: Chemometrics & Informatics
thumbnail image: Multidimensional algorithm: Cancer detector

Algorithmic assessment

Gordon Okimoto. Credit: UH Cancer Center

A new algorithm that can mine data from tumour samples more effectively than before could improve diagnostics of liver, ovarian and other cancers, according to research from the University of Hawaii Cancer Center.

Gordon Okimoto, Ashkan Zeinalzadeh, Tom Wenska, Michael Loomis, James Nation, Tiphaine Fabre, Maarit Tiirikainen, Brenda Hernandez, Owen Chan, Linda Wong, and Sandi Kwee report a computational algorithm to analyze so-called "big data" obtained from various tumour samples. "A growing problem in cancer research is figuring out how to analyse the many kinds of big genomic data for different cancers. The overwhelming quantity and complexity of the data has created an analysis bottleneck that has slowed the translation of the knowledge within the data to the clinic," explains Okimoto, who is Co-Director of Biostatistics and Informatics Shared Resource. "We have figured out a way to mine these data for the benefit of cancer patients," he adds.


The team's algorithm - Joint Analysis of Many Matrices by ITeration (JAMMIT) - can identify different patterns across multiple molecular data types such as gene expression and genetic mutations that when taken together accurately predict what treatments would be best for a specific cancer patient. The algorithm could thus speed up the approval of novel treatments for various cancers as well as helping to improve prognosis. In addition, in these times of economic strife, it could reduce the cost of treating cancer. Moreover, the algorithmic approach could open up new ways to carry out more precisely managed clinical trials.

The initial data processed and extracted by JAMMIT is on liver and ovarian cancer from private research laboratories and public databases such as The Cancer Genome Atlas (TCGA), a data archive of the US National Cancer Institute. The analysis identified small sets of genes that accurately predict which patients would benefit most from chemotherapy. These same signatures also suggest that many of the ovarian and liver cancer patients studied would benefit from combining chemotherapy with immunotherapy.

"Sparse matrix approximations of rank-1 provide a simple yet effective means of jointly reducing multiple, big data types to a small subset of variables that characterize important clinical and/or biological attributes of the bio-samples from which the data were acquired," the team reports.

Bio mining

The liver cancer results, specifically, were based in part on tissue samples collected locally by the liver cancer working group (TeamLiver). TeamLiver includes about 15 researchers and physicians from the Cancer Center and local hospitals that make up the Hawaii Cancer Consortium collaborating on liver cancer research for more than four years. Hawaii has one of the highest rates of liver cancer in the nation and the second highest liver cancer mortality rate in the USA.

Of course, Okimoto and colleagues now plan on extending their analyses to data from dozens of other types of cancer. Okimoto and Thomas Wenska have also started SNR Analytics, Inc a company focused on securing the intellectual property and requisite funding for the development of a computational pipeline based on the JAMMIT algorithm for the discovery of predictive gene signatures for cancer and other complex diseases.

"We show that JAMMIT outperforms other joint analysis algorithms in the detection of multiple signatures embedded in simulated MDDS. On real multimodal data for ovarian and liver cancer we show that JAMMIT identified multi-modal signatures that were clinically informative and enriched for cancer-related biology," the researchers report in the journal BioData Mining.

"We plan on analyzing multi-modal data for all 30-plus cancers in The Cancer Genome Atlas (TCGA) using an analysis pipeline based on the JAMMIT algorithm," Okimoto told SpectroscopyNOW. "This will first require 'industrializing' the JAMMIT pipeline as a cloud-based application that is robust, and easy to access and use." He suggests that this will then accelerate the extraction and modelling of clinically relevant information contained in public repositories such as TCGA and molecular, imaging, and clinical data acquired by private labs.

"One of our goals is to identify low-dimensional molecular signatures that form the basis for data-driven models that accurately predict the clinical trajectory of cancer for an individual patient," he adds. Those models would then help clinicians to take a more personal and precise approach to treating cancer as well as allowing researchers to identify gene networks that can be targeted to better treat or perhaps even cure cancer. "Our signatures identify ovarian and liver cancer patients who would derive a significant survival benefit by combining standard chemotherapy and immunotherapy," Okimoto offers as an example. Fundamentally, he hopes "to accelerate the translation of knowledge contained in big biomedical data sets to the clinic using methods from computational sciences."

Related Links

BioData Mining 2016, 9,online: "Joint analysis of multiple high-dimensional data types using sparse matrix approximations of rank-1 with applications to ovarian and liver cancer."

Article by David Bradley

The views represented in this article are solely those of the author and do not necessarily represent those of John Wiley and Sons, Ltd.

Follow us on Twitter!

Social Links

Share This Links

Bookmark and Share


Suppliers Selection
Societies Selection

Banner Ad

Click here to see
all job opportunities

Most Viewed

Copyright Information

Interested in separation science? Visit our sister site

Copyright © 2019 John Wiley & Sons, Inc. All Rights Reserved