Notes on genome-grist style metatranscriptomics analysis

[taylorreiter]

• Quantify against CDS seqeunces
  • reasoning: microbes express polycistronic transcripts, but genomic rearrangements and other biology between strains makes it hard to have good references at the operon level.
  • Modified code for CDS download from genbank if it exists, or genome download + pyrodigal CDS prediction if it does not exist: https://github.com/taylorreiter/2022-microberna/blob/main/best_genome_reference.snakefile#L189 (i didn't actually run this code inside my snakefile, so there's a non-zero chance it won't work. but I think it's a good template/example)
• Quantify reads using salmon in single end mode, regardless of whether the library is pe or se.
  • Citation in isolates: https://www.biorxiv.org/content/10.1101/2022.01.24.477642v1
  • reasoning: microbes express polycistronic transcripts. If we quantify against CDS sequences, relationships between genes in the same operon are not preserved, so many pe reads will discordantly map to two CDS sequences that are part of the same operon. I believe salmon discards discordantly mapped reads, which is likely a lot of data, and will be biased between different organisms based on the number and size of operons in the genome
• Consider providing the genome or intergenic regions as decoys to the salmon transcriptome to label reads that map to the intergenic regions of polycistronic transcripts (for operon discovery)
  • sorry if that bullet sounds incomplete, i'm still testing that strategy and don't want to dump a ton of explanation into it until I have good results

I'll probably have more thoughts/ideas/lessons in the next couple months, but there are my takes right now

[taylorreiter]

Oh also tRNAs, rRNAs, and ncRNAs sometimes recruit like 30% of reads. Not sure if those sequences are in the CDS files, but they could be cut out of the genome from a gff annotation file, or using bakta (derivative of prokka that I've started to really like) to do annotation.

[falimaucdavis]

Thanks for the suggestions, Taylor. I really appreciate it. Regarding this metatranscriptomic data, we are working on analysis steps that we put together when we met with Titus. I anticipate that likely we will run into some roadblocks and we might bug a little bit with that if it is OK. Besides the metatranscriptomic dataset, Hugo and I would love to meet with you at your convenience to chat a bit more about our current approach for 16S amplicon studies that were used for a few projects. If could let us know the best way to schedule a meeting we appreciate it. Thanks in advance for the opportunity to collaborate with you. All the best. Fabio