Cancer classification: MALDI imaging distinguishes different types of cancer

Skip to Navigation

Ezine

  • Published: Apr 1, 2012
  • Author: Steve Down
  • Channels: Proteomics & Genomics / Proteomics
thumbnail image: Cancer classification: MALDI imaging distinguishes different types of cancer

Cancer discrimination

Six common types of tumour have been differentiated by mass spectrometry in a proof-of-principle proteomics study by European scientists, which could lead to a personalised treatment regime for cancer sufferers.

Cancer diagnosis is a tricky business, especially for those tumours that have spread to other parts of the body. These metastatic cancers, as they are known, can spring up in distant organs and account for a major proportion of all diagnosed cancers. On a global scale, nearly 13 million cancer cases were diagnosed in 2008, according to the latest figures from the WHO.

In 2.3-4.2% of all cases worldwide, metastatic cancers are diagnosed as a cancer of unknown primary when the original cancer type cannot be identified. This imposes a detrimental effect on the treatment regime, which depends upon tumour classification, and highlights the desire for accurate cancer classification.

A team of scientists working in Germany and Switzerland has taken a large step towards this goal by employing mass spectrometry in a new proteomics study. It was described by Axel Walch and colleagues from the German Research Center for Environmental Health (Helmholtz Zentrum Munchen), Neuherberg, the Technical University of Munich, the University of Heidelberg, and the Department of General, Visceral and Vascular Surgery, Baden Hospital.

There have been various reports highlighting the ability of proteomics to distinguish healthy tissue from cancerous tissue. In general, they were achieved using samples of homogenised tissue but this is not ideal because the results must be interpreted under the assumption that the tissue is homogeneous. In addition, the required sample sizes can exceed the amounts typically harvested during diagnosis.

The solution reported by Walch was to use matrix-assisted laser desorption/ionisation (MALDI) imaging, in which thin tissue sections typically 12 µm thick were analysed. Small samples like this are easily sectioned from biopsy samples and the slices retain the tissue morphology. After MALDI imaging, the same sections can be stained for histological comparison, so cancerous and non-cancerous areas can be distinguished.

 

The magic of MALDI imaging

In initial experiments, a total of 171 tissue samples from six types of cancer were sectioned and prepared for MALDI imaging. Barrett's cancer, breast cancer, colon cancer, hepatocellular carcinoma, gastric cancer and thyroid carcinoma were represented. The spectra were collected by scanning across the tissue slices to produce a series of peaks corresponding to the proteins and peptides within the tissue. Only those spectra originating from cancerous tissue, selected by comparing with the stained slices, were extracted from the data and inspected.

The protein complements of each cancer type were sufficiently divergent so that even visual examination of the spectra could distinguish between them. However, the team employed a statistical analysis based on the R software package for accurate comparison. Initially, two-thirds of the spectra of each sample were randomly selected to make up a training set and the remainder were placed in a test set. The m/z values were selected and submitted to two classifiers within the R software - the Support Vector Machine (SVM) and the Random Forest (RF).

Both classifiers yielded almost perfect tumour classification within the training set, with high accuracy (82.7 and 81.2%, respectively) for the test set, the comparable figures indicating that the techniques are robust. The misclassifications were attributed to subtypes within particular cancers. For instance, breast cancer is known to have five distinct subtypes with different molecular features.

There have been a few published reports of tumour distinction, although not for six different types, but this new work goes further by providing a training set for validating the accuracy of the classifiers. To illustrate this, Walch and the team added colon cancer liver metastasis (CCLM) into the samples and ran the two classifiers again.

Both systems discriminated the liver metastasis samples from primary colon cancer and hepatocellular carcinoma with an accuracy greater than 80%. In addition, the CCLM samples were correctly classified (>99% in the training set and >71% in the test set) when colon cancer primary tumours were used to create the training set.

The specificities were very high, although the sensitivities were relatively low. However, these would be improved by introducing distant metastases of known origin into the training set, to give more accurate classifications.

So, it appears that MALDI imaging provides a good basis for tumour classification. "It might open new fields in tissue sample classification. This proof-of-principle study shows for the first time that proteomic classification of solid tumour entities can be highly accurate while needing a minimal amount of tissue," said Walch.

Successful application of this technique in the clinical setting will aid the design of appropriate personalised treatment regimes for patients that are directed towards the specific cancer that should give a greater likelihood of a successful outcome.

Related Links

Journal of Proteome Research 2012, 11, 1996-2003: "Tumor classification of six common cancer types based on proteomic profiling by MALDI imaging"

Article by Steve Down

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

Social Links

Share This Links

Bookmark and Share

Microsites

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 separationsNOW.com

Copyright © 2017 John Wiley & Sons, Inc. All Rights Reserved