Normalize Parameters

Table of Contents


What is normalization?

In a typical analysis, researchers have control samples which indicate the expected behavior under known conditions. A typical negative control may be the growth of an isolate on blank media (i.e. low or no growth) and a typical positive control may be the growth on a preferred susbtrate (e.g. glucose).

When testing the growth of an isolate on a new condition, it often helps to compare it to other known conditions especially negative or positive controls. AMiGA has the normalization function which simplifies this process for the user. Normalization can be performed by either subtracting the the growth paameters of control sample(s) from the treatment samples, or by dividing the growth parameters of the treatment samples from the control sample(s).

Normalization by subtration will result in values that are centered around zero which would indicate no difference from the control samples, for example,

\[\text{Normalized AUC} = \text{AUC}(\text{Isolate on Substrate}) - \text{AUC}(\text{Isolate on Blank Media})\]

while normalization by dividion will result in values that are centered around 1 which would indicate no difference fom the control samples, for example,

\[\text{Normalized AUC} = \frac{\text{AUC}(\text{Isolate on Substrate})}{\text{AUC}(\text{Isolate on Blank Media})}\]


How to normalize growth parameters?

After modelling growth curves and estimating growth parameters using the fit function, users can adjust the estimated growth parameters with the normalize function. To properly normalize data, the users must indicate in their meta-data how samples are grouped and which samples are considered the control samples in each group. For example, all 96 wells belong to the same group because they are inoculated by the same cell culture and the A1 well is the Negative Control well. As previously shown, AMiGA automatically adds a Group and Control column in the meta-data for Biolog PM plates to streamline this normalization.

Example 1

As an example, let us assume that we profiled an isolate with two replicate plates of Biolog PM1

python $amiga/amiga.py fit -i /Users/firasmidani/experiment/ -o split_merged


This would have generated the following split_merged_summary.txt table

Well_ID Plate_ID Isolate Substrate Group Control auc_lin … t_gr
A1 erand_PM1-1 erand None 1 1 0.21 … 4.4
B1 erand_PM1-1 erand L-Arabinose 1 0 0.85 … 4.1
C1 erand_PM1-1 erand N-Acetyl-D-Gluosamine 1 0 0.83 … 4.0
… … … … … … … … …
F12 erand_PM1-1 erand Inosine 1 0 0.23 … 4.5
G12 erand_PM1-1 erand L-Malic Acid 1 0 0.33 … 3.8
H12 erand_PM1-1 erand 2-Aminoethanol 1 0 0.68 … 4.0
… … … … … … … … …
A1 erand_PM1-2 erand None 2 1 0.22 … 4.5
B1 erand_PM1-2 erand L-Arabinose 2 0 0.85 … 4.0
C1 erand_PM1-2 erand N-Acetyl-D-Gluosamine 2 0 0.83 … 3.9
… … … … … … … … …
F12 erand_PM1-2 erand Inosine 2 0 0.19 … 4.6
G12 erand_PM1-2 erand L-Malic Acid 2 0 0.29 … 3.9
H12 erand_PM1-2 erand 2-Aminoethanol 2 0 0.72 … 4.0


We can adjust the growth paarmeters by simply using the normalize function.

python $amiga/amiga.py normalize -i /Users/firasmidani/experiment/summary/split_merged -o "new_normalized" --normalize-method "division"
  • -i points to the summary file generated by the fit function.
  • -o allows the user to pick the name for the new summary table. If the user does not pass the -o arugment, AMiGA will create a new file with a _normalized suffix, for example, split_merged_normalized.txt. If the user wants to replace the file with a new version that includes normalized parameters, they must add the --over-write argument.
  • --normalize-method specifies that the user wants to adjust the growth parameters by division. The user user can instead opt for subtraction which is the default option.


However, you can also run the above command more explicitly if you did not have the Group and Control columns in the data.

python $amiga/amiga.py normalize  -i /Users/firasmidani/experiment/summary/split_merged --over-write --group-by "Plate_ID" --normalize-by "Substrate:None"


Example 2

Let’s say you have a simple experiment with two isolates on two conditions (minimal media or trehalose). Below we show the mapping file that we created and we have already fit the curves for which the summary file was saved as CD89_wt_summary.txt.

  Plate_ID Isolate Substrate Ribotype Comments Group Control
A1 CD_treA CD89_wt Minimal Media RT027 Wild-type 1 1
A2 CD_treA CD89_wt Minimal Media RT027 Wild-type 1 1
A3 CD_treA CD89_wt D-Trehalose RT027 Wild-type 1 0
A4 CD_treA CD89_wt D-Trehalose RT027 Wild-type 1 0
B1 CD_treA CD89_ko Minimal Media RT027 treA knock-out 2 1
B2 CD_treA CD89_ko Minimal Media RT027 treA knock-out 2 1
B3 CD_treA CD89_ko D-Trehalose RT027 treA knock-out 2 0
B4 CD_treA CD89_ko D-Trehalose RT027 treA knock-out 2 0


We can adjust the growth paarmeters by simply using the normalize function.

python $amiga/amiga.py normalize -i /Users/firasmidani/experiment/summary/CD89_wt_summary.txt -o "CD89_wt_summary_normalized" --normalize-method "division"


If we did not expliclty include the Group and Control columns, we can still normalize the data directly with the command.

python $amiga/amiga.py normalize -i /Users/firasmidani/experiment/summary/CD89_wt_summary.txt -o "CD89_wt_summary_normalized" --normalize-method "division" --group-by 'Isolate' --normalize-by 'Substrate:Minimal Media'


How to normalize parameters from a summary of pooled replicates?

In this example, I have in my working directory (/Users/firasmidani/experiment/) multiple data files in the data sub-folder . Let us assume that I fit growth curves for replicates pooled across all unique combinations of Isolate and Substrate.

python $amiga/amiga.py fit -i /Users/firasmidani/experiment/ -o "pooled_analysis" --pool-by "Isolate,Substrate"


I can specify that AMiGA should normalize each group (unique combination of Isolate and Substrate) by their wells where the Substrate is also Negative Control. In this case, each unique model is specified by an Isolate and a Substrate. If there are multiple control wells, AMiGA will normalize to the mean of their growth parameters.

python $amiga/amiga.py normalize -i /Users/firasmidani/experiment/summary/pooled_normalized_merged_by_ribotype -o "pooled_normalized_merged_by_ribotype_normalized" --group-by "Isolate,Substrate" --normalize-by "Substrate:Negative Control"


Command-Line arguments

To see the full list of arguments that amiga normalize will accept, run

python normalize.py --help

which will return the following message

usage: amiga.py [-h] -i INPUT [--over-write] [--verbose] [--group-by GROUP_BY]
                [--normalize-by NORMALIZE_BY]
                [--normalize-method {division,subtraction}]

Compare two growth curves

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
  --over-write          over-write file otherwise a new copy is made with
                        "_normalize" suffix
  --verbose
  --group-by GROUP_BY
  --normalize-by NORMALIZE_BY
  --normalize-method {division,subtraction}


See more details for these arguments in Command Line Interface