I'm currently working on a genome liftover task and need some advice on the most optimal method for generating chain files, particularly when aligning genomes of the same species. I’ve come across a few options and would appreciate hearing from those who have experience with them.
- Preprocessing: faToTwoBit, fasplit, and Genome Fragmentation:
I've seen suggestions to preprocess genomes using tools like faToTwoBit or fasplit to convert genome files to the 2bit format or break them into smaller fragments. Does anyone have insights into whether this preprocessing improves chain generation? Specifically, does the conversion to 2bit format optimize the process, or could it potentially complicate things? Also, does fragmenting the genome prior to alignment have any advantages or disadvantages when generating chain files?
- BLAT vs Minimap2 for Alignments
Both BLAT and Minimap2 are popular choices for genome alignments, but I’m unsure which one would be better for closely related species. BLAT seems good for "tight" alignments, but I’ve heard Minimap2 is preferred for larger, more complex genomes. Has anyone had experience with these aligners for chain file generation in the case of closely related species? Which one works better in this context, and why?
- MAF2Chain vs axtChain
Once alignments are done, converting them into chain files is the next step. I've seen that MAF2Chain is a valid option, but axtChain is often cited as the standard for generating high-quality chain files. Is MAF2Chain a viable alternative, or does axtChain remain the preferred tool, especially when dealing with genomes of the same species? Any insights on how these two tools compare?
- Preprocessing Considerations
From what I’ve gathered, preprocessing steps such as filtering repetitive blocks with ChainAntiRepeat or netting the chains with chainNet can significantly improve chain quality. Does anyone have experience with these preprocessing tools and can confirm their impact on chain file generation?
Looking forward to your thoughts and experiences. I’m hoping to optimize this process for a genome liftover task, and any suggestions or guidance would be greatly appreciated!
Thanks in advance!
I necessarily can't speak to what are the best options, but it could be worth trying out some easy to run pipelines like LiftOn (https://github.com/Kuanhao-Chao/LiftOn) and flo (https://github.com/wurmlab/flo). once you have some results, you can then consider customizations and tweaks to make things better