Phaedra 2 – Concepts

Documentation: Screening

Mon, 01 Jan 0001 00:00:00 +0000

Screening is obviously a very broad area of scientific research. In this document, we limit the scope to screening of plate-based carriers using digital microscopy. This means that Phaedra deals with image and/or numeric data generated by taking readouts of plates using a digital microscopy instrument.

Plate

Plates are physical containers that have a number of wells embedded in them. Common formats are 8x12, 16x24 and 32x48 wells per plate. Phaedra supports arbitrary dimensions, provided that the plate has a rectangular format.

Substances of interest are added into the wells, and the activity of the substances is measured either once or repeatedly by having an instrument perform a readout of the plate.

Experiment

An experiment in Phaedra is simply a collection of plates. Grouping plates correctly into experiments is important for some functionality such as experiment-wide normalization, plate polishing or multiplo dose-response curve fitting, where data from multiple plates is taken together for calculation. In other cases, experiments are simply a means for organizing your data efficiently.

Measurement

When an instrument takes a readout of a plate, the resulting dataset is called a measurement. In Phaedra, measurements may contain up to three categories of data:

Image data: images taken of each well, possibly across multiple channels (wavelengths). If multiple regions (fields) of a well are imaged, they will be stitched together by Phaedra into a single image per well per channel. In addition, if the image analysis routine generated outline masks or overlays, those may be captured into the measurement as well.
Subwell data: a numerical 2D dataset for each well where the rows represent “subwell items”, such as single cells. An example is the single-cell dataset generated by an image analysis routine.
Well data: a numerical 2D dataset for the whole plate where the rows represent wells.

Substances and Compounds

A common use case for Phaedra is to process data from small molecule or compound screening. Here, compounds are inserted into plate wells and their effects are measured by an instrument. When such a plate is defined in Phaedra, each well may be annotated with one or more substances such as compounds. A substance in Phaedra is always defined by a substance type, name and a concentration.

Documentation: Data Model

Mon, 01 Jan 0001 00:00:00 +0000

Projects, Experiments and Plates

The core entity of Phaedra, the plate, is stored in a container entity called an experiment. Experiments are in turn grouped into projects. While experiments and projects can be thought of as folders, plates are data entities with a lot of information attached to them.

Nearly all of the calculations and operations that happen in Phaedra, are centered around plates.

Generally speaking, plates require two pieces of data to be attached to them, before they can be fully functional:

A plate definition, such as a linked plate layout template
At least one measurement dataset containing well data

In addition, you can add more data to make the plate more feature-rich and to give you a more complete insight into the plate’s data:

Well Images for one or multiple channels/stainings, generated by the instrument
Well Image overlays, such as cell segmentation overlays or masks generated by the image analysis
SubWell data, e.g. single-cell data, kinetic or time-lapse data, typically generated by the image analysis
Tags and properties, to facilitate searching through many plates or to annotate them for custom processing

There is no fixed order in which this data should be added to a plate: you can start by defining some blank plates and update them as more data becomes available. The calculation and other data processing steps are typically automated via a pipeline and will execute as soon as their required pieces of data become available.

Experiments are often used to represent one batch or one run of plate measurements. You can freely move plates between experiments (keeping in mind that functionality like multiplo curve-fitting will trigger on every change to the experiments’ contents), and experiments may grow to contain dozens or even hundreds of plates (e.g. for High-Throughput screening).

Projects can also be as small or as large as you like. They can run for years, accumulating hundreds of experiments. The main point about projects is that they are configured with a default pipeline, which tells Phaedra which automation steps to perform on which conditions. This means that any plate that is entered into the project, automatically becomes subject to execution by the pipeline.

Plate Layout Templates

As mentioned in the previous topic, one of the essential bits of information for a plate is its plate definition. The plate definition informs us of the layout of the plate’s contents, such as compound samples or control wells. More precisely, a plate definition gives, for each well in the plate:

The well type describing the role or purpose of the well in the assay. Commonly used well types are: SAMPLE, LOW CONTROL, HIGH CONTROL, EMPTY, etc.
For SAMPLE wells, the type, name and concentration of the substance that is, or will be, inserted into the well.
Optionally one or more annotations that can be used in further calculations.

Plate Measurements

Whereas the plate definition provides information about the plate’s contents, which is usually available well before any readout of the plate takes place, the plate measurement contains data that is obtained during the readout. In many cases, this measurement data consists of “raw” data, such as TIF image files and instrument configuration files. A primary analysis is performed on this raw data, generating additional data such as numeric data tables.

It’s important to note that Phaedra does not perform primary analysis: this is typically done by the software that is bundled with the instrument. Instead, Phaedra will take in any raw data (both images and numeric tables) and work from there to process the data and produce downstream results.

Documentation: Calculation

Mon, 01 Jan 0001 00:00:00 +0000

Protocols, Features and Formulas

In terms of data calculation, the feature is the central concept that Phaedra uses. Features are grouped into protocols which can be executed on plates to produce result data.

When a protocol is executed on a plate (a process called plate calculation in Phaedra), a result dataset will be generated yielding one numerical value per well per feature. So if a plate contains 384 wells and a protocol with 5 features is executed on it, you will end up with a resultset containing 384 x 5 = 1920 feature values.

The calculation formula is the most important setting of a feature. This formula may reference one or more columns from the plate measurement on which the calculation will take place. Alternatively, a feature formula may also reference the output of another feature, effectively creating a chain of calculations. For example, a formula named “ratio A/B” might reference two columns A and B from the measurement, and calculate the ratio between them.

Protocols are versioned, and any change to the protocol (or one of its features) will cause the version number to increase. This allows you to see a precise trail for any resultset, showing you which version of which protocol was used on which plate measurement to obtain the data you see in the resultset.

Pipelines

While it is perfectly possible to manually import measurements, link them to plates, execute protocols and generate reports by hand, this can be a very time-consuming process, especially when working on large projects and dealing with large volumes of plates.

Instead, you can set up pipelines which represent sequences of actions that will be performed automatically on any plate that comes along. All pipeline executions are tracked, so you can see if any plate failed to process due to missing or malformed data, for example.

Pipelines are defined by a sequence of steps, usually starting with a DataCapture step. Steps can be refined with triggers, specifying the conditions that must be met before the pipeline step will be executed.

For example, suppose that you have two projects, A and B. You want each project to have its own pipeline, so that they can each perform project-specific steps; perhaps in project A you’ll want to generate a QC report for each finished experiment and in the other you don’t. Therefore, you define two pipelines: pipelineWithReport and pipelineWithoutReport. In pipelineWithReport, you configure the first step with a pattern so that it only processes measurements that have “projectA” in their name. Accordingly, you configure a pattern in pipelineWithoutReport to match all measurements with “projectB” in their name.

Documentation: The User Interface

Mon, 01 Jan 0001 00:00:00 +0000

Phaedra’s user interface is customizable, although the overall layout is fixed:

In the top is the title bar containing the Main Menu button and an icon representing the currently logged-in user.
On the left is the Navigator, showing a number of categories you can open and navigate into.
In the center of the workbench is the main area, where detailed information is shown for the entity you’ve navigated to.
On the right is the side panel, which is collapsed by default. This panel can be expanded to display additional views on the current entity.

The title bar cannot be moved, but the three main sections (left, center, right) below it can be resized by clicking and dragging the separator bars between them.

Sections

When displaying larger amounts of information, the main area is usually split up into multiple sections. These sections can be recognized by their blue color and the little white chevron on the right. Clicking on the chevron will collapse or expand the whole section.

Bookmarks

Since the Phaedra workbench is entirely browser-based, you can easily bookmark specific locations. For example, if you use the Navigator to browse into your projects and then open “Project X” which has ID 123, you’ll note that the browser’s address bar shows something like this:

https://your.phaedra.server/project/123

Opening this URL, e.g. by clicking on it from a bookmark tab or by receiving it from a colleague by mail, will immediately open a Phaedra workbench displaying “Project X”.

Note: obviously, the person opening the link may receive a login screen first to confirm their access before being able to see the project!