How to deal with sequences of unequal length?

[amin-sagar]

How to deal with sequences of unequal length?

[josephsbrown1]

Hi Amin, thanks for reaching out and I am so glad you are applying this work. Yes, it is excellent to drop duplicates. In the data curation steps, there is a drop duplicates step, but it could be good to add to the Make Map by Dim Reduction. I am going to have a set of revisions to this GitHub soon and will include that.

As for the length, yes the peptide sequence are best to be the same length, and that error you see does seem to be from it. I will also get a different error if the library contains sequences of a different length, usually this:

ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (2,) + inhomogeneous part.
If the peptides are not the same length, the UMAP has a tendency to separate the peptides based on length. The way that I saw this was by "zero-filling" the encoded vector of short peptides during the encoding step to match the length of the longest peptide. It of course would be useful to have this option available because it may work for others and I will consider adding it as an option for users in the next update due in March.

Thank you again,
Best,
Joe

[amin-sagar]

Thanks @josephsbrown1

It would be great to have the zero filling feature. My peptides differ only by a couple of residues in length and there are some semi-conserved motifs. It would be interesting to see if those can be picked up despite the variations in length.
I look forward to the next update.

Meanwhile, I am trying to add a step where the input dataframe is split into multiple dataframes based on the length of the peptides and the analysis is done on each subset separately while ignoring any subsets if the number of peptides is below a certain number.

Best,
Amin.