I would use the raw data, not the mean.

1. Compute the empirical cumulative distribution function (ecdf) of each series

2. Use the Kolmogorov-Smirnoff test to assess the difference between the two ecdf's

You could try a Tanimoto similarty.

https://en.wikipedia.org/wiki/Jaccard_index#Tanimoto.27s_definitions_of_similarity_and_distance

They are identical on appearance and the SD overlap mostly. You could compare residuals or errors :)

I'd be curious to compare the 3 methods (and I would provide the comparison). Would you please post the raw data?

Hi Everyone, Thanks for the suggestions.

I'm looking into ECDFs and Tanimoto similarities, still trying to work out how to practically apply those to the data. Regarding the residuals and errors, without fitting the data to a model I'm not sure how I could get those.

Renaud, thanks for offering. I've attached the raw data.

Here is the implementation (with R) of the ECD idea.

Thanks for sharing the raw data.

Here is a small perl script to compute the Tanimoto coefficient.

As input two text file are needed, each with just one data column of the y-axis measurement. I shifted the data by +10 to get only positive values. Then the Tanimoto is between 0 and 1. The higher would mean the more similar one.

The values are, however, all quite close to each other.

Renaud, thanks so much for the help. This is a really interesting way to compare the data.

Ardim, thanks for the script. I don't have experience with perl, but it is well annotated. I will try to get it running.

I have updated the doc with the computation of the Tanimoto coefficient as well as a simulation test to assess its significance.

Hi, Here are a few papers from the Biopharma industry which you may find interesting:

Bierau, Horst, et al. "Higher-order structure comparison of proteins derived from different clones or processes." BioProcess Int 6.8 (2008): 52-9.

Teska, Brandon M., et al. "Comparison of quantitative spectral similarity analysis methods for protein higher-order structure confirmation." Analytical biochemistry 434.1 (2013): 153-165.

Dinh, Nikita N., et al. "Quantitative spectral comparison by weighted spectral difference for protein higher order structure confirmation." Analytical biochemistry 464 (2014): 60-62.

The issue of establishing similarity through biophysical characterization is key to the biosimilars industry and the FDA and EMA are starting to demand appropriate statistics to ensure objective assessments of risk. These concepts are also important in setting batch release criteria for biotherapeutics in general.

I spend a lot of my time supporting our customers in these areas so would be happy to provide further information if it would be helpful.

Cheers - Doug