Home

Welcome to the genppi wiki!

Source code

As promised, considering we finally publish the article, we open-sourced genppi's code.

Predicting for low amino acid identity proteins

Our software poses challenges even for ourselves, the creators. There is no manual about how to accomplish a perfect interaction network. I needed to generate an interaction network comprising 224 genomes of the Desulfobacterales Order. We were not able to create a trustable network, a network with an unbiased degree distribution, using the GENPPI default parameters. The GENPPI compiled versions are limited to the parameters -aadifflimit 0 -aacheckminlimit 18. However, we must relax the amino acid identity with -aadifflimit 3 to create interactions among the alpha, beta, and gamma Coenzyme-B sulfoethylthiotransferase subunits. The only way to achieve such a task is by modifying the code from:

                 (setf aadifflimit  (if parametervalue
                                       (if (numberp (read-from-string parametervalue))
                                         (if (<= (parse-integer parametervalue) 1) 
                                           (parse-integer parametervalue)

To

                 (setf aadifflimit  (if parametervalue
                                       (if (numberp (read-from-string parametervalue))
                                         (if (<= (parse-integer parametervalue) 3) 
                                           (parse-integer parametervalue)

There is a strong motivation for applying such a restriction in the source code. The potential pan-genome size is enormous for -aadifflimit greater than 1. It is not practical to run GENPPI with 224 genomes and use the -aadifflimit 3 parameter. Our workaround was to choose two close related genomes, including the specific one for further analysis, and predict an interaction network for this pair using -aadifflimit 3.

genppi-lowidentity -aadifflimit 3 -aacheckminlimit 18  -dir pair/

We also ran the analysis of the 224 genomes.

genppi -expt dynamic -ws 20 -ppcomplete -ppdifftolerated 1 -ppaadifflimit 1\ 
                   -aadifflimit 0 -aacheckminlimit 18 -dir full/

To create a final interaction network, we blended the interaction networks of our target genome from both folders (pair and full), resulting in a not biased degree distribution. There were about 12 thousand out of 112 thousand edges in common between both webs. However, GEPHI joined them, summing the weights.

[Anderson Rodrigues dos Santos]

I fixed a bug in GENPPI when the parameter -pphistofilter was omitted. That was the case when, for instance, no parameters were passed but the working directory. This new version implements parallelism in the central bottleneck: the pan-genome generation. Running times decreased to one-third in a 40 cores machine compared to the mono-processed version.