Skip to content

🐟 🍣 🍱 Highly-accurate & wicked fast transcript-level quantification from RNA-seq reads using selective alignment

License

Notifications You must be signed in to change notification settings

sk7-dotcom/salmon

Β 
Β 

Repository files navigation

salmon logo

Build Status Documentation Status install with bioconda

Try out alevin (salmon's single-cell processing module)! Get started with the tutorial

Try out the new alevin-fry framework for single-cell analysis!

Help guide the development of Salmon, take our survey

Pre-computed decoy transcriptomes

tl;dr: fast is good but fast and accurate is better ! Although the precomputed decoys (<=v.14.2) are still compatible with the latest major release (v1.0.0). We recommend updating your index using the full genome, as it can give significantly higher accuracy. For more information, please check our extensive benchmarking comparing different alignment methods and their performance on RNA-seq quantification in the latest revised preprint manuscript. Please use the tutorial for a step-by-step guide on how to efficiently index the reference transcriptome and genome for accurate gentrome based RNA-seq quantification.

Specifically, there are 3 possible ways in which the salmon index can be created:

  • cDNA-only index : salmon_index - https://combine-lab.github.io/salmon/getting_started/. This method will result in the smallest index and require the least resources to build, but will be the most prone to possible spurious alignments.

  • SA mashmap index: salmon_partial_sa_index - (regions of genome that have high sequence similarity to the transcriptome) - Details can be found in this README and using this script. While running mashmap can require considerable resources, the resulting decoy files are fairly small. This will result in an index bigger than the cDNA-only index, but still mucch smaller than the full genome index below. It will confer many, though not all, of the benefits of using the entire genome as a decoy sequence.

  • SAF genome index: salmon_sa_index - (the full genome is used as decoy) - The tutorial for creating such an index can be found here. This will result in the largest index, but likely does the best job in avoiding spurious alignments to annotated transcripts.

Facing problems with Indexing ?, Check if anyone else already had this problem in the issues section or fill the index generation request form

What is Salmon?

Salmon is a wicked-fast program to produce a highly-accurate, transcript-level quantification estimates from RNA-seq data. Salmon achieves its accuracy and speed via a number of different innovations, including the use of selective-alignment (accurate but fast-to-compute proxies for traditional read alignments), and massively-parallel stochastic collapsed variational inference. The result is a versatile tool that fits nicely into many different pipelines. For example, you can choose to make use of our selective-alignment algorithm by providing Salmon with raw sequencing reads, or, if it is more convenient, you can provide Salmon with regular alignments (e.g. an unsorted BAM file with alignments to the transcriptome produced with your favorite aligner), and it will use the same wicked-fast, state-of-the-art inference algorithm to estimate transcript-level abundances for your experiment.

Give salmon a try! You can find the latest binary releases here.

The current version number of the master branch of Salmon can be found here

Documentation

The documentation for Salmon is available on ReadTheDocs, check it out here.

Salmon is, and will continue to be, freely and actively supported on a best-effort basis. If you need industrial-grade technical support, please consider the options at oceangenomics.com/support.

Chat live about Salmon

You can chat with the Salmon developers and other users via Gitter!

Join the chat at https://gitter.im/COMBINE-lab/salmon

About

🐟 🍣 🍱 Highly-accurate & wicked fast transcript-level quantification from RNA-seq reads using selective alignment

Resources

License

Code of conduct

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 97.6%
  • C 1.5%
  • CMake 0.6%
  • Shell 0.2%
  • Python 0.1%
  • Dockerfile 0.0%