Although the concept of autophagy — the process for degrading and recycling cellular components — has been around since the 1960s, a deeper understanding of this cellular mechanism wasn’t realized until Yoshinori Ohsumi began experimenting with yeast to identify the genes involved in autophagy, beginning in the 1990s. His discoveries won him the 2016 Nobel Prize in Physiology or Medicine, announced this past October.
Ohsumi’s work “led to a new paradigm in our understanding of how the cell recycles its content,” the Nobel Assembly at Karolinska Institutet, which awards the prize, announced. “His discoveries open the path to understanding the fundamental importance of autophagy in many physiological processes.”
Disrupted or mutated autophagy has been linked to Parkinson’s disease, Type 2 diabetes, cancer, and more. This has major implications for studying human health and developing new strategies for treating disease.
Breaking the Bottleneck: The Role of HCS in Drug Discovery Research
Autophagy is best understood by applying high content screening (HCS) - also referred to as high content analysis (HCA) - analytical approaches. HCS/HCA are sophisticated image-analysis and computational tools that have become useful in breaking the industrial biomolecular screening bottleneck in drug discovery research. The bottleneck stems from image processing, statistical analysis of multiparametric data, and phenotypic profiling – at both the individual cell and aggregated well level.
HCS tools, described as “essentially high-speed automated microscopes with associated automated image analysis and storage capabilities”, are necessary — as high throughput screens — to identify cells, recognize features of interest, and tabulate those features.
It is no small task. To validate and automate a phenotypic HCS analysis requires:
- • data management
- • image processing
- • multivariate statistical analysis
- • machine learning (based on hit selection)
- • profiling at the individual and aggregated well-level
- • decision support for hit selection
HCS in Phenotypic Profiling of Autophagy
We validated a phenotypic image and data analysis workflow using an autophagy assay. Secondary analysis provided an automated workflow for extensive data visualization and cell classifications that furthered understanding of the multiparametric phenotypic screening data sets from three different cell lines.
An end-to-end solution that includes reagents, instruments, image-analysis tools, and informatics facilitates screening breakthroughs and successful screening experiments.
The solution should serve both small-scale experiments and analysis of a small number of plates and automated methods for larger data sets.
Inspection of the data can be done using an unsupervised machine learning technique called Self-Organizing Map algorithm – a type of artificial neural network that clusters similar profiled data points together. In our autophagy assay of the HeLa, HCT 116, and PANC-1 cells, this analysis showed that the phenotypic response to chloroquine is very different in the three cell lines – something almost indistinguishable by eye.
Semi-supervised machine learning methods were used to perform Feature Selection and Hit Classification. By reducing the number of parameters to just the most relevant ones, hit stratification and classification enabled identification of which wells are ‘autophagosome positive’ or ‘autophagosome negative.’
Both supervised and semi-supervised machine learning led to similar EC50 curves in the autophagy assay. The supervised linear classifier is helpful whenever the phenotypes are predicted by the user, while unsupervised classification might be better suited for applications with an unknown number of phenotypes. Read the full application note here.
While the underlying pathways still need further research for complete understanding, we are able today to merge phenotype and genotype to help measure autophagy quantitatively in different cell lines, using end-to-end HCS solutions.
Serving industry with HCS solutions for more than a decade, PerkinElmer offers a number of integrated solutions to fully enable autophagy analyses at the speeds required for today’s drug discovery.
Interested in incorporating autophagy into your research, and leveraging the available HCS technologies that can analyze, classify, and interpret phenotypic screening data to deliver fast insight?
Check out this webinar to learn about an autophagy assay across three cancer cell lines. This assay was done to validate and automate a phenotypic HCS analysis workflow using PerkinElmer’s Columbus and High Content Profiler, powered by TIBCO Spotfire.