How to Go from Cell Image to Insight, with Multivariate Analysis

Biologists spend a lot of time observing cells under the microscope and making qualitative assessments. Do they look right? Are they dividing? What’s the differentiation status? These observations inform many important downstream decisions. But the hardest question remains: how can we get quantitative insights from complex cell morphology data?

^{This article is posted on our Science Snippets Blog} 

Why Quantify Cell Morphology

Cell morphology is one of the strongest indicators of cell health. But understanding it can be complicated as cells change their morphology for a variety of reasons. Let’s say your cells are shrinking. Is it because they are dividing, dying, or simply running out of space?

Quantifying cell morphology over time helps with quality control both to identify cross-contamination of cell types, and to ensure that cell lines maintain consistent properties over multiple rounds of subculture. In this context, cell morphology can change for a variety of reasons:

• Crowding: With increasing confluence, some cells become smaller and more homogeneous in size and shape.
• Environmental/chemical: External triggers can cause blebbing, rounding or other unusual appearances. 
• Cell death: Regardless of the initial morphology, dying cells become small, circular, and highly textured.

Advanced Tools Empower Cell Analysis 

Developments in cell imaging and computational power have transformed how scientists analyze cells. Unlike traditional microscopy methods that are cumbersome and slow, advanced systems allow direct visualization and quantitative tracking of multiple parameters related to cell health, morphology and function.

One example are the Incucyte^® Live-Cell Analysis Systems. Each system incorporates a microscope that sits inside a tissue culture incubator. Without ever moving the cells, you can acquire HD phase and fluorescence images of live cells and automatically segment and analyze individual cells using software tools.
 

Deriving Insights Using Multivariate Analysis

With today’s instrument technologies it’s easy to amass a lot of imaging data quickly. The next step is translating complex datasets into meaningful quantitative information that help scientists make decisions. 

Software tools are increasingly bridging the gap between multi-dimensional data and actionable insights. For example, metrics relating to cell shape, such as area, texture, roundness, and perimeter length can be analyzed individually (univariate analysis) or using a combination of metrics (multivariate analysis).

Solutions like the SIMCA^® software package have tools for multivariate data analysis, which is used to explore relationships between variables, or factors, in complex datasets. 
 

Analyzing Cell Morphology Over Time

This example demonstrates how two such advanced tools can be combined to perform multivariate analysis in one workflow. 

Here, cell data was first acquired on an Incucyte^® Live-Cell Analysis System and segmented using the integrated software. Next, segmentation metrics were entered into SIMCA^® for principal component analysis, which is a type of multivariate analysis, that simplifies high-dimensional data and clusters data based on patterns in an unbiased way. 

This analysis represents A549 cells treated with compounds that either arrest cell growth (cycloheximidex), or cause cell death (camptothecin). The PCA plot visualizes the change in morphology over time:

• Arrested cells (pink), which do not increase in confluence, show very little change over time.
• Dying cells (teal) show rapid change and increased activity from pro-death caspase enzymes, as expected.
• Healthy cells (grey), which are actively growing, show more change over the same time period.

^{Footnote for image: Quantification of morphology in healthy cells (vehicle control, VEH, grey), arrested cells (cycloheximide, 1 µM CHX, pink) and dying cells (camptothecin, 10 µM CMP, teal) using principal component analysis and compared on a PCA score plot (A). Time is indicated by circle size where the smallest circle corresponds to 0h, and largest to 72h. Overlay of vehicle cell confluence and change in PC1 over time (B). Overlay of PC1 and caspase activation (green fluorescence intensity) in CMP-treated cells (C).}

The ability to quantify total cell morphology from microscopy images is a huge untapped resource. Combining live-cell imaging with multivariate analysis has created a powerful workflow for analyzing total cell morphology through time.