Photolithographic Tumor DNA Isolation

点击翻译按钮获取中文摘要

Abstract: Photolithographic Tumor DNA Isolation Personalized oncology is based on the idea that the mutations in a cancer determine optimal therapy. Mutation detection is increasingly possible with newer next generation sequencers and bioinformatics, but currently the first step of DNA isolation is ad hoc, manual, and non-standardized. Human tumors are complex mixtures of normal and tumor cells, and mutation detection would become more reliable and reproducible if nearly pure tumor DNA was extracted. Here we propose to develop an automated system that can extract >90% pure tumor DNA from conventional H&E stained microscope slides by integrating photolithography with high resolution slide scanners, image analysis algorithms, and modern 3D printers. Machine learning image algorithms can distinguish tumor from normal cells, and this information will be transferred to the 3D printer, which places opaque material directly over tumor nuclei on the slide. The slide is then exposed to short wave UV light to destroy the DNA in unprotected normal cells whereas tumor DNA is selectively protected by the photolithographic mask. DNA can be extracted from the entire slide, and only DNA in the protected tumor cells can be sequenced (whole exome or targeted sequencing) or measured for CpG methylation. The spot resolution of the 3D printer is about 40 microns, and therefore very irregular complex topography and small features like a single tumor gland can be protected by photolithography. The entire system (scan, analyze, print, irradiate, extract) could yield >90% pure tumor DNA from an H&E slide in about 24 hours. The transformational potential is that the system converts a currently ad hoc, highly labor-intensive technical step into an automated, well-documented and reproducible extraction that can add information and learning because the exact extracted cell numbers, their phenotypes and spatial locations are known. Because it uses an image algorithm to select the tumor cells, the “same” DNA isolation can be performed by anyone anywhere in the world. Moreover, image algorithms can “learn” to better extract tumor DNA based on feedback from the DNA sequencing. The integration of this system into a sequencing pipeline would improve the reliability, documentation, and reproducibility of mutation calling by extracting nearly pure (>90%) tumor DNA, which will advance both reproducible cancer research and the clinical translation of precision oncology.