71 / 129
Mechanobiology of Disease
Poster Abstracts
66
67-POS
Board 67
Nuclear Mechanical Biomarkers for Cancer Diagnosis
Aneesh R. Sathe
1
, Karthik Damodaran
1
, Caroline Uhler
2
, Shivashankar G.V.
1,3
.
1
National University of Singapore, Singapore, Singapore,
2
Massachusetts Institute of
Technology, Cambridge, MA, USA,
3
Italian Foundation for Cancer Research (FIRC), Milan,
Italy.
Current diagnosis for cancer employ a number of nuclear morphometric measures. While these
approaches have provided insights into late stage disease diagnosis, early diagnosis has been a
major challenge. Picking up subtle changes in nuclear morphometrics at an early stage is
important for better prognosis and therapeutic intervention. In this study we use single cell
imaging methods combined with machine learning to quantitatively detect nuclear morphometric
features between normal and cancer cell lines. High throughput widefield images of DAPI
stained nuclei are used to extract around 2000 physical and texture features as biomarkers.
Linear discriminant analysis (LDA) of these biomarkers was able to discriminate between
normal and breast cancer cell lines. In addition, LDA analysis discriminated fibrocystic (MCF
10A) and metastatic (MCF7 and MDA-MB-231) human breast cancer cell lines with high
accuracy. Further we describe a compressive loading assay of single cells to amplify the
differentiability of these cell types based on structural memory in chromatin condensation states.
Finally using sparse LDA we identify the most predictive biomarkers which can be used as
diagnostic reporters of altered nuclear architecture. Our methodologies collectively provide a
simple and robust platform for potential applications in early disease diagnostics.