Integrative Prediction Model Uses “Omics”, Histopathology to Improve Cancer Prognosis
News Nov 29, 2017 | Original Story by Krista Conger for Stanford Medicine.
Consider it a pre-Thanksgiving brain teaser. The image at the right is something called an autostereogram — a three-dimensional picture that can be seen only by staring through and beyond the colored, seemingly random pattern. (You may need to make it bigger, or even to zoom in.) What do you see? Go ahead, I’ll wait.
The picture is of an abstract smiley face in a square frame. Once you see it, it seems crystal clear. But look away and the image is hidden again.
If you were able to make it work, it is because your brain successfully overlapped the two sides of the pattern to visualize a previously hidden image. Former Stanford graduate student Kun-Hsing Yu, MD, PhD and geneticist Michael Snyder, PhD, used a similar layering technique to help researchers and clinicians analyze patches of cancer cells from tissue biopsies. They’ve worked out a way to overlay the visual information on the slide — the shapes and positions of cells — with another layer of information about the cells’ ‘omics’ — a term coined by Snyder to encompass the DNA sequence (the genome), the RNA molecules (transcriptome) and even the proteins (the proteome) present in individual cells.
The research appears in Cell Systems.
Yu explained "Histopathology evaluation, the visual assessment of microscopic patterns of tumor tissue, is the gold standard for diagnosing many types of cancers. However, the molecular processes underpinning the morphological changes observed in cancer cells are largely unknown. In addition, the clinical utility of integrating histopathology slides and omics data had not been explored. In this study, we successfully revealed the gene and protein expression patterns in lung cancer tissues and showed that dysregulated genes associated with cell de-differentiation [an early step that can lead to cancer] participated in cell-cycle, DNA replication, and p53 signaling pathways.
By layering these multiple types of information, the researchers were able to identify morphological patterns associated with mutation of the p53 protein, which is an important prognostic factor for lung cancer patients. When they integrated this information with other omics data they were able to generate personalized prognoses for individual patients.
As Yu explained "Our integrative prediction model can better inform oncologists of their patients’ prognoses than histopathology, clinical history or omics information alone, and could allow clinicians to better formulate treatment plans based on a patient’s predicted survival. It can also help to classify patients into disease subtypes when histopathlogy results are ambiguous. In short, the systematic incorporation of the ‘big data’ of omics research into medical practice can assist in personalized cancer treatment as well as facilitate the development of precision medicine."
This article has been republished from materials provided by Stanford Medicine. Note: material may have been edited for length and content. For further information, please contact the cited source.
Yu KH, Berry GJ, Rubin DL, Ré C, Altman RB, Snyder M. Association of Omics Features with Histopathology Patterns in Lung Adenocarcinoma. Cell Syst. 2017 Nov 13. pii: S2405-4712(17)30484-2. doi: 10.1016/j.cels.2017.10.014. [Epub ahead of print] PubMed PMID: 29153840.
Machine learning – a field of artificial intelligence that uses statistical techniques to enable computer systems to ‘learn’ from data – can be used to analyse electronic health records and predict the risk of emergency hospital admissions, a new study from The George Institute for Global Health at the University of Oxford has found.
International Conference on Clinical Trials & Pharmacovigilance
Feb 28 - Mar 01, 2019