Closed holgerbrandl closed 6 years ago
I'd be still curious to learn more about bootstrapping support in kallisto, so feel welcome if you need more info to process my request.
Gene level estimates are not stored in the hdf5 files, but are computed with sleuth, https://github.com/pachterlab/sleuth
The user group, https://groups.google.com/forum/#!forum/kallisto-sleuth-users is a better place to get help on using the sleuth functions. Also take a look at http://pachterlab.github.io/sleuth/ and the sleuth walk-throughs to see how sleuth is used.
Ok, thanks. Actually I was briefly looking into the sleuth-repo already but its REAMDE stated that
It makes use of quantification uncertainty estimates obtained via kallisto for accurate differential analysis of isoforms
That's why I ended up here.
@holgerbrandl: Both can be correct as you asked for gene level estimates. IRC, kallisto
estimates transcript (isoform) level uncertanties, which are read by sleuth
and summarized to the gene-level.
My bad, I actually wanted to to access the "transcript level uncertainties". Where are they stored? I somehow can not find them when reading in h5 or tsv output from kallisto.
The h5 files contain the complete quantifications of each bootstrap. If you want summaries I recommend using sleuth to compute them, take a look at get_bootstrap_summary
in sleuth, which should give you what you need.
Cool, thanks (also for your patience), that's the answer I was looking for.
I managed to run kallisto with bootstrapping. Now I'm struggling to get some summary statistics. I can read the raw bootstrap results (e.g. in R) with
I've tried
tximport
package already but could not find a corresponding method.However, I'd be rather curious to get gene-level summary estimates about the uncertainty of expression estimates. Is this part of the hdf5 or do I need to calculate those numbers myself? If the latter applies, are there some public code-bits I could reuse?