example, tenfold cross-validation indicates the sample data is
split into ten groups. One group in turn is used as a test dataset
and the remaining groups used for training. The average of the
k evaluation scores provides an indication of how the model is
expected to perform when used to make predictions on data
not used during model training.
15. The distributed version of Vacceed is configured to run ML
algorithms via R functions contained in packages. The algorithms are executed using Rscript. There are three R functions
in install_dir/vacceed//evidence that encapsulate the relevant command for each algorithm:
_wrapper.R, _runPred.R, and _makePred.R,
where is the algorithm abbreviation. Parameters to finetune the algorithms can be modified in _makePred.R
(e.g., parameters “ntree” and/or “mtry” in rf_makePred.R,
where rf ¼ random forest, ntree ¼ number of decision trees,
and “mtry” ¼ number of variables to try at each split in the
decision tree).
16. Run the following command to see available class I alleles:
./src/predict_binding.py IEDB_recommended mhc
(only listed alleles can be used).
17. Run the following command to see available class II alleles:
python mhc_II_binding.py allele (only listed alleles can be
used).
18. May need to append new program location to the PATH
variable.
19. This is a template script only and will need to be edited appropriately to suit the new program. There are user comments
denoted by a “#” symbol, but a familiarity with Linux scripting
is expected.
20. Amending get_evidence.pl requires experience in writing Perl
scripts. Reading step 8 under the section “Adding a new
resource” in the Vacceed User Guide may prove useful when
amending get_evidence.pl.
Acknowledgments
SJG gratefully acknowledges Zoetis (Pfizer) Animal Health for
funding the development of Vacceed through a PhD scholarship.
Eukaryotic Pathogen Antigen Discovery Using Vacceed
41
split into ten groups. One group in turn is used as a test dataset
and the remaining groups used for training. The average of the
k evaluation scores provides an indication of how the model is
expected to perform when used to make predictions on data
not used during model training.
15. The distributed version of Vacceed is configured to run ML
algorithms via R functions contained in packages. The algorithms are executed using Rscript. There are three R functions
in install_dir/vacceed/
where
(e.g., parameters “ntree” and/or “mtry” in rf_makePred.R,
where rf ¼ random forest, ntree ¼ number of decision trees,
and “mtry” ¼ number of variables to try at each split in the
decision tree).
16. Run the following command to see available class I alleles:
./src/predict_binding.py IEDB_recommended mhc
(only listed alleles can be used).
17. Run the following command to see available class II alleles:
python mhc_II_binding.py allele (only listed alleles can be
used).
18. May need to append new program location to the PATH
variable.
19. This is a template script only and will need to be edited appropriately to suit the new program. There are user comments
denoted by a “#” symbol, but a familiarity with Linux scripting
is expected.
20. Amending get_evidence.pl requires experience in writing Perl
scripts. Reading step 8 under the section “Adding a new
resource” in the Vacceed User Guide may prove useful when
amending get_evidence.pl.
Acknowledgments
SJG gratefully acknowledges Zoetis (Pfizer) Animal Health for
funding the development of Vacceed through a PhD scholarship.
Eukaryotic Pathogen Antigen Discovery Using Vacceed
41