A question about the example.

[lllastronaut]

Hello, you have done a great job and thank you for your contribution to the field of protein-based genome representation.

When I was embedding the genome, I used the sample you provided (test.faa) and used the command (esm-embed --input ./test1.faa --outdir ./test_output --esm esm2_t6_8M --torch-hub test) to convert the protein sequence into esm2 embedding, and then used the command (pst graphify --file ./test_output/esm2_t6_8M_results.h5 --fasta-file ./test.faa --output ./test_graph.h5) to try to convert the esm2 embedding into a graph structure. An error occurred when converting the graph structure:
###################################################
ValueError: FASTA file headers must be in prodigal format: '>scaffold_ptn#' with the additional metadata separated by ' # '
####################################################
It seems to be a problem with the input file test.faa of the parameter --fasta-file ./test.faa. You mentioned in the Readme and -h that it should conform to the prodigal format (>scaffold_ptn#). I tried to modify the name line (the line starting with >) in test.faa to the prodigal format (>SAMEA.110_1 or >SAMEA_1), but the above error still occurred. How can I solve this problem? I noticed that you didn't provide a sample file for the complete process. If my understanding is wrong, could you provide the detailed format of the prodigal, or provide all the sample files required for the complete process? Thank you.

[cody-mar10]

Hi, thanks for reaching out.

I am updating the wiki to better explain all of this and will provide an example FASTA file to run the end-to-end pipeline.

The protein FASTA headers should look like this:

>SAMEA2737773_b1_ct14_vs2@Podoviridae__sp._ctp7i14@linear_1 # 143 # 1474 # -1 # ID=1_1;partial=00;start_type=ATG;rbs_motif=AGxAGG/AGGxGG;rbs_spacer=5-10bp;gc_cont=0.371

which is the standard output from prodigal/pyrodigal. The field separator is #, and the relevant fields for the graph format are the 1st (protein name and position in the scaffold) and 4th (encoding strand).

It sounds like your protein FASTA files do not follow the prodigal format, which is fine, but the strand information is still needed. You can create your own tab-delimited file for the encoding strand that looks like this:

genome1_1   1
genome1_2   1
genome1_3   1
genome1_4   -1

The protein names are in the first column (should be identical to the names in the FASTA file) and the strand in the second column. If you do not know what strands your proteins are encoded on, it should be ok to set them all to the same strand, but I haven't tested what that does in terms of genome representations.