STAR VS bowtie2

I will be using this PC to run Analyses using R studio and Python. For the most part I will be working with genomic sequencing data and other cell data. I will be using different bash software (STAR, FeatureCounts) and other packages in R (Seurat) and Python ((UMAP)[https://umap-learn.readthedocs.io/en/latest/]). I also wanted to try using some other analyses software such as 10x Genomics' CellRanger.

According to this, the issue should have been resolved in 2.7.4a. Does anyone know how I could go about fixing this? Appreciate any thoughts.

The main claim is that pseudo-alignment is much faster than traditional alignment, but often I see pseudo-alignment to transcriptome compared to alignment to genome. Though STAR at least has issues building indices for transcriptomes. I wanted to try to benchmark myself if alignment to transcriptome is much slower than pseudo-alignment, but it asked for 100gb of RAM to create an index for 250mb of mm10 transcriptome. This issue shows that STAR shockingly requests for 1TB of ram to build an index for a 900mb axolotl transcriptome. So even if the alignment itself might be fast, the indexing is not fit-for-purpose.

Thanks for this. I ended up installing windows subsystem for linux (WSL). Could I ask if you've used STAR on WSL (ubuntu) before? I'm having trouble with the steps after compiling the STAR executable, which I was able to:

I believe the most straightforward solution would be to add a tag that designates which gene the read overlaps (if any). Lucky for you, this is now implemented in star as of version 2.7.3a. I've actually not used this feature yet, but I believe it is done using the flag --outSAMattributes GN see here.

2) Use bowtie2 to align reads against CHM13. This will let you separate human from nonhuman (important, as human sequences are a common contaminant in many nonhuman genomes).