Making quickmodel more flexible to accept different plate layouts

[MehtapIsik]

At the moment quickmodel is designed to work this plate layout:

• odd rows: protein + ligand1
• even rows: buffer + ligand1

I would like to analyse an experiment with a different layout:
Row A : protein1 + ligand1
Row B : protein2 + ligand1
Row C : protein3 + ligand1
Row D : protein4 + ligand1
Row E : protein5 + ligand1
Row F : protein6 + ligand1
Row G : buffer + ligand1
Row H : buffer + ligand1

I will try to modify quickmodel script to handle this case. Can I add it as an alternative quickmodel script to assaytools for now? @sonyahanson

I think this is the part I need to modify:
https://github.com/choderalab/assaytools/blob/master/scripts/quickmodel.py#L74-L75

[MehtapIsik]

@sonyahanson You can find my modified quickmodel script for this new layout:
quickmodel_layout2.py
I changed lines 73-75 of quickmodel.py.

I used the default inputs.py format but repeated the name of the ligand 4 times.

[MehtapIsik]

Can we define the assay layout using a dictionary like fashion with assay_format, protein_order (this will replace the protein section) and ligand_order sections in inputs.py.

assay_format can be row-wise or column-wise

It will be useful to have assay_format that defines in which direction each titration series lie, so we can analyse data in different orders and even when the experiments are designed in columns (Column 1 - protein, column 2 buffer etc.)

For example our usual layout would be defined like this:

'assay_format' : 'row-wise' 
'ligand_order' :  {'A': 'Bosutinib', 'B':'Bosutinib Isomer', 'C':'Erlotinib', 'D':'Gefitinib', 'E':'Bosutinib', 'F': 'Bosutinib Isomer', 'G': 'Erlotinib', 'H': 'Gefitinib'}
'protein_order' : {'A': 'p38, 'B':'buffer, 'C':'p38', 'D':'buffer', 'E':'p38', 'F': 'buffer', 'G': 'p38', 'H': 'buffer'}

And an unusual layout like I did with HSA samples will be defined like this:

'assay_format' : 'row-wise' 
'ligand_order' :  {'A': 'dansylamide', 'B':'dansylamide', 'C':'dansylamide', 'D':'dansylamide', 'E':'dansylamide, 'F': 'dansylamide', 'G': 'dansylamide', 'H': 'dansylamide'}
'protein_order' : {'A': 'HSA1, 'B':'HSA2', 'C':'HSA3', 'D':'HSA4', 'E':'HSA5', 'F': 'HSA6', 'G': 'buffer', 'H': 'buffer'}

An example for column-wise experiment where only first 4 columns are used:

'assay_format' : 'column-wise' 
'ligand_order' :  {'1': 'Bosutinib', '2':'Bosutinib', '3':'Gefitinib', '4':'Gefitinib''}
'protein_order' : {'1': 'buffer', '2':'Src', '3':'buffer', '4':'Src'}

[sonyahanson]

I agree it would be good to have a more flexible plate layout! I think that if we want it this general, it might be good to start the transition to autoprotocol for the initial file parsing: https://github.com/choderalab/assaytools/blob/master/examples/autoprotocol/README.md

We can have our data in a format that's appropriate for the more advanced analysis, while still doing the analysis the same way we are, and then we don't have to manage too many things at once.

[sonyahanson]

Perhaps as an intermediate, we can create a section in the 'inputs.py' where we can define assay_format as alternating_row or block_row, which seem to be the two formats we have been using recently.