phraber / lassie Goto Github PK

Computing TF loss may be more efficient in the seqinr::consensus method.

After installing lassie with devtools::install_github("phraber/lassie") I found that logo-related things like lassie::make.timepoint.logos(eg.swarmset) failed with a permission denied message. Apparently the zip file that devtools downloads from https://api.github.com/repos/ doesn't have the weblogo Python script marked executable, even though it is executable here in the repository.

Thanks for lassie! I was working through the paper and vignette and found that a few items weren't working for me, so I've made a fork with some possible solutions. I'll submit issues and matching pull requests, so you can have a look and see if you'd like to merge them in.

One problem I had was that plot(eg.swarmset) failed with this error:

Error in paste(c(env, shQuote(command), args), collapse = " ") : 
  promise already under evaluation: recursive default argument reference or earlier problems?

I traced that down to make.logoplot's color_option argument not being specified. It looked to me like the simplest fix was just to provide the option like make.timepoint.logos currently does.

To sort samples by time-point sequenced, assume the labels are sorted lexicographically.
That is, d10 will be sorted before d5 unless the labels use leading 0s, e.g. d05 and d10.
The sample timepoint is a character value rather than numeric because it is used internally for column naming.

vignette('lassie') suggests this should work, though it does not. Need to update vignette.

a <- swarmtools(aas_file=alignment_file, tf_loss_cutoff=0:5*20)

Warning message:
In set.tf.loss.cutoff(retval, tf_loss_cutoff) :
ERROR in set.tf.loss.cutoff(): Please provide a single numeric value.

Need to automate installation of weblogo and ensure it is accessible on arbitrary system configurations.

One more for you - we've found that in alignments where all amino acids at a given site are the same across timepoints but different from the transmitted founder, report.variant.frequencies fails. I think I have a fix so I'll put the details in another pull request.

The lassie vignette shows this for a swarmset example:

eg.swarmset <- swarmset(eg.swarmtools, is_verbose=FALSE)

But the is_verbose argument was removed in 2c94f44.