Editing Nanopore RNA Sequencing Protocol (section)

====EpiNano-Error (Pair-wise)====

If we have two samples, one that is '''unmodified''' and one that is '''modified''' (in <code>Rscript Epinano_DiffErr.R ...</code>, the '''ko.csv''' is the unmodified sample), we can compare them using the ''R script'' <code>Epinano_DiffErr.R</code>. This script fits a linear regression model between paired samples, then determines outliers that often are due to base modifications.

<code>Bash</code>
<syntaxhighlight lang="Bash">
SAMPLEID_UNM="FAR91556" \
SAMPLEID_MOD="FAM95931" \

Rscript $EPI/Epinano_DiffErr.R \
-k $NALI/${SAMPLEID_UNM}/${SAMPLEID_UNM}_${CHR}.plus_strand.per.site.csv \
-w $NALI/${SAMPLEID_MOD}/${SAMPLEID_MOD}_${CHR}.plus_strand.per.site.csv \
-t 3 \
-o ${SAMPLEID_UNM}_vs_${SAMPLEID_MOD}_${CHR}  \
-c 30 -f sum_err -d 0.1 -p
</syntaxhighlight>

In the above code, <code>-t</code> indicates the minimum z-score threshold (i.e., standard deviations from the mean) you would like to reach for a mod to be detected. <code>-o</code> indicates the name prefix of the output file you wish to generate from the run. <code>-f</code> indicates the feature (via column name) used to predict the modification. <code>-d</code> indicates the minimum deviance required of the selected feature between the two samples. -p is included here, which generates EpiNano plots, but you can omit that and run Epinano_Plot.R instead if you wish.

Lastly, I’ll want to move these files to the EpiNano output folder. We can do that with this block:

<code>Bash</code>
<syntaxhighlight lang="Bash">
ETARGET="$EPIOUT/${SAMPLEID_UNM}_vs_${SAMPLEID_MOD}" \
mkdir $ETARGET

mv ${SAMPLEID_UNM}_vs_${SAMPLEID_MOD}_${CHR}.delta-sum_err.prediction.csv \
${SAMPLEID_UNM}_vs_${SAMPLEID_MOD}_${CHR}.linear-regression.prediction.csv \
${CHR}.${SAMPLEID_UNM}_vs_${SAMPLEID_MOD}_${CHR}.delta-sum_err.bar.pdf \
${CHR}.${SAMPLEID_UNM}_vs_${SAMPLEID_MOD}_${CHR}.linear-regression.scatter.pdf $ETARGET
</syntaxhighlight>

We will now move forward and I will show you how to do the alternative and run single samples with ''EpiNano-SVM''.