AAindex has 566 indices with real values for each of the 20 amino acids mostly from wetlab measurements. Many of these values are highly correlated, and thus do not make good features for classification algorithms, due to redundancy/duplication. However, in the past, I had read that they can be clustered to a non-redundant (uncorrelated) set of around to to 70 indices.

I cannot find any list of non-redundant/uncorrelated entries in the AA index database.

Where can I find such a list?

I am not particular interested in making one myself, since I know this has already been done and validated. I just cannot find the reference...

  • $\begingroup$ Could you please clarify in which format do you have such data? Also if you indicate what is this data perhaps we can help indicating how to cluster or remove some data points. Did you look at the literature referenced in the home page of the AAindex? What is your goal? Why do you want to reduce it to 70 and not the typical 20? $\endgroup$ – llrs Sep 10 '18 at 15:21
  • $\begingroup$ @Llopis The AA index has a list of 566 different measures for each of the 20 amino acids; i.e. 11320 values. This is too many to deal with when resources are limited. AA index is basically a flat database with 566 rows which are correlated, which can be flattened to around 70 uncorrelated rows. $\endgroup$ – Aalawlx Sep 10 '18 at 15:28
  • 1
    $\begingroup$ Then you might be interested in the principal component analysis method or the multidimensional scaling method. But it is difficult to answer, you seem to have in mind a paper or a method and the information you provide (at least to me) is not enough to identify it $\endgroup$ – llrs Sep 10 '18 at 15:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.