nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

DoE Examples on Microtiter Plates/Well-Plates

Steph_Georges

Hello SE Community,

A customer Champion is interested in organizing a discovery session on Design of Experiments (DoE) with JMP. However, to make the session relevant, he would like all DoE examples to be based on microtiter plates (96WP) to properly address the associated challenges of his company.

1. Examples

Would you have any JMP examples to share on this theme?

2. Specific Demo Items

Regarding their specific challenges, their Champion would like to see the following cases:
a) Accounting for row/column effects – Their robot has a constraint, and apparently, some factors need to remain constant within a row or a column.
b) Incorporating an additional temperature effect per the plate.

Tell me if I am wrong, but for 2a, I think it would be possible to address this challenge like this

Capture d'écran 2025-02-17 155741.png

As JMP does not have a "Very Very Hard to Change" option, how can the temperature effect be implemented in these conditions?

My non-expert opinion is as follows: if there are three temperature conditions to study (Min, Middle, Max), I would quadruple the initial DoE using four plates at the following temperatures: Middle, Min, Max, Middle. The additional Middle condition would help separate the plate effect from the temperature effect.

What is your opinion on this approach? Is there a cleaner way to handle it?

Looking forward to your examples and insights!

Stéphane

Accepted answers

Byron_JMP

Since we are working with whole plates here. Temperature can be a fixed effect, however the whole plate would need to be repeated at each temperature.

In microtiter plates, often several columns are reserved for controls or standard curves.
Often each of the 8 rows are part of a dilution series. 1, .5, .25, .125... this is why so many 4p curve fits have 8 points. Because of this, each column on the plate is the experimental unit, not each well. Dilution is not part of the DOE experimental design.

The first step in the analysis is fitting curves to the data in each column on the plate. Then the parameters (like LD50, or IC50) become the responses of the DOE.

Split Plot.jmp

Phil_Kay

Hi @Steph_Georges ,

Some posts in the TRL that might be useful:

Using Fit Curve and Curve DOE for Cell Growth Analysis - System Engineers - Confluence

Well Plates - Microtiter Plates Up to 9600 - System Engineers - Confluence

I'm also attaching a paper from AstraZeneca that has some nice case studies of big biological experiments using well plates.

And shameless plug for my podcast, "The Next Experiment", available on YouTube, Apple and Spotify. It should be relevant for this audience.

Cheers,
Phil

Pharmaceutical Statistics - 2024 - Jones - Quality by Design for Preclinical In Vitro Assay Development.pdf

Byron_JMP

In the attached table use this JSL to open the excel file. It looks super impressive if you use the excel import wizard.

Once it's open, you can fit 4p curves to the data using treatment as the grouping variable.

Open(
	"$Examples/microtiter plate kinetic experiment.xlsx",
	Worksheets( "96 Well Kinetic Data" ),
	Use for all sheets( 1 ),
	Concatenate Worksheets( 0 ),
	Create Concatenation Column( 0 ),
	Worksheet Settings(
		1,
		Has Column Headers( 1 ),
		Number of Rows in Headers( 2 ),
		Headers Start on Row( 22 ),
		Data Starts on Row( 24 ),
		Data Starts on Column( 1 ),
		Data Ends on Row( 0 ),
		Data Ends on Column( 0 ),
		Replicated Spanned Rows( 1 ),
		Replicated Spanned Headers( 0 ),
		Suppress Hidden Rows( 1 ),
		Suppress Hidden Columns( 1 ),
		Suppress Empty Columns( 1 ),
		Treat as Hierarchy( 1 ),
		Multiple Series Stack( 0 ),
		Import Cell Colors( 0 ),
		Limit Column Detect( 0 ),
		Column Separator String( "-" )
	)
);

microtiter plate kinetic experiment.xlsx

russ_wolfinger

Once the data are collected from the DoE, mixed models are recommended and powerful, and provide natural leverage for JMP Pro. Plus they are fun to set up--happy to help out.

Florian_Vogt

Hi @Steph_Georges ,

I had been asked about this topic before. Particularly, we were challenged by competitor solutions: Modde offers what's called RED-MUP designs. After some back and forth with internal experts we found there is more modern ways to set up well plate designs, but unfortunately this is what the customers were used to do for years...

I created three different scenarios to build such designs (see the attached journal). I am adding some notes around RED-MUP I wrote back then when dealing with the customer.

Description

Rectangular Experimental Designs for Multi-Unit Platforms (RED-MUPs)are designs made for experiments with multi-channel pipettes on well plates e.g. 96 well plate.

The RED-MUPs are built from two separate design protocols: an 8-experiment design (Des_8) with c factors and another 12-experiment design (Des_12) with r factors. The Des_8 design is laid out on to the plate so that every row has the same combination of values, while the Des_12 design is added so that each column consists of the same combination of values (Fig. 1).

Open image-20240202-135621.png

Stock solutions are made according to these two designs and are added to the plate with multi-channel pipettes. Hence all possible combinations of the stock-solutions are made and an overlaid design with c+r factors results (Fig. 1). Typically, in the RED-MUP designs, more wells (read: experiments) are available than in classical designs. Hence, the goal is to streamline the work within the actual number of available experiments.

Using an eight-by-twelve 96-well plate it is possible to investigate up to five factors (2+3) in an RSM design, or up to eighteen (7+11) factors in a screening design. Various eight and twelve experiment RSM precursor-designs and the resulting RED-MUP protocols were thoroughly investigated.

When planning a RED-MUP protocol it is important to have a clear understanding of the investigated assay or system. In particular, some factors may be linked together, and thus for practical reasons have to be part of the same precursor design. Since interactions can be resolved in screening designs, i.e. Res III + Res III = Res III/ IV, it is also worthwhile to consider potential interactions that warrant investigation and to position the factors involved in these putative interactions in separate precursor designs.

Situation(s)

Why did it end up here? How did it end up here? What is the practical situation?

The topic was raised by Syngenta, who are doing lots of experiments by high throughput screening. Basically, it is required to do DoE for HTS experiments, but setting up the DoE’s is not easily done.

Modde PRO can apparently create RED-MUP designs for up to 4 plates with sizes 8x12 and 16x24, with 32x48 size plate.

JMP does not offer a built in solution for RED-MUP. This is causing people use Modde over JMP for DoE in particular departments.

All comments

Byron_JMP

Since we are working with whole plates here. Temperature can be a fixed effect, however the whole plate would need to be repeated at each temperature.

The first step in the analysis is fitting curves to the data in each column on the plate. Then the parameters (like LD50, or IC50) become the responses of the DOE.

Split Plot.jmp

Phil_Kay

Hi @Steph_Georges ,

Some posts in the TRL that might be useful:

Using Fit Curve and Curve DOE for Cell Growth Analysis - System Engineers - Confluence

Well Plates - Microtiter Plates Up to 9600 - System Engineers - Confluence

I'm also attaching a paper from AstraZeneca that has some nice case studies of big biological experiments using well plates.

And shameless plug for my podcast, "The Next Experiment", available on YouTube, Apple and Spotify. It should be relevant for this audience.

Cheers,
Phil

Pharmaceutical Statistics - 2024 - Jones - Quality by Design for Preclinical In Vitro Assay Development.pdf

Byron_JMP

In the attached table use this JSL to open the excel file. It looks super impressive if you use the excel import wizard.

Once it's open, you can fit 4p curves to the data using treatment as the grouping variable.

Open(
	"$Examples/microtiter plate kinetic experiment.xlsx",
	Worksheets( "96 Well Kinetic Data" ),
	Use for all sheets( 1 ),
	Concatenate Worksheets( 0 ),
	Create Concatenation Column( 0 ),
	Worksheet Settings(
		1,
		Has Column Headers( 1 ),
		Number of Rows in Headers( 2 ),
		Headers Start on Row( 22 ),
		Data Starts on Row( 24 ),
		Data Starts on Column( 1 ),
		Data Ends on Row( 0 ),
		Data Ends on Column( 0 ),
		Replicated Spanned Rows( 1 ),
		Replicated Spanned Headers( 0 ),
		Suppress Hidden Rows( 1 ),
		Suppress Hidden Columns( 1 ),
		Suppress Empty Columns( 1 ),
		Treat as Hierarchy( 1 ),
		Multiple Series Stack( 0 ),
		Import Cell Colors( 0 ),
		Limit Column Detect( 0 ),
		Column Separator String( "-" )
	)
);

microtiter plate kinetic experiment.xlsx

russ_wolfinger

Once the data are collected from the DoE, mixed models are recommended and powerful, and provide natural leverage for JMP Pro. Plus they are fun to set up--happy to help out.

Florian_Vogt

Hi @Steph_Georges ,

I created three different scenarios to build such designs (see the attached journal). I am adding some notes around RED-MUP I wrote back then when dealing with the customer.

Description

Rectangular Experimental Designs for Multi-Unit Platforms (RED-MUPs)are designs made for experiments with multi-channel pipettes on well plates e.g. 96 well plate.

Open image-20240202-135621.png

Situation(s)

Why did it end up here? How did it end up here? What is the practical situation?

Modde PRO can apparently create RED-MUP designs for up to 4 plates with sizes 8x12 and 16x24, with 32x48 size plate.

JMP does not offer a built in solution for RED-MUP. This is causing people use Modde over JMP for DoE in particular departments.