(2) multiply each coverage value by the appropriate normalization factor. In the resulting file, all converted libraries are
corrected for differences in sequencing depth relative to each
other, and all nonconverted libraries are corrected for differences in sequencing depth relative to each other. Differences in
depth between converted and nonconverted libraries are more
nuanced and will be addressed in the next step.
3.11 Filter to Obtain
Above-Background 5
0
Ends and Ensure
Correct Normalization
Between Converted
and NonConverted
Libraries
1. To identify 5
0 ends that are most likely to be biologically
meaningful, we suggest three filters (see Fig. 2):
(a) Remove 5
0 ends with normalized coverages of 5 reads
or less.
(b) Retain only the 5
0 end with the highest coverage in a 5-nt
window. This is useful because transcription initiation is
imprecise, and as a result, there will be substantial numbers of 5
0 ends mapping within a few nt both upstream and
downstream of the primary TSS (see Note 15).
(c) Retain only 5
0 ends that meet a read depth threshold
relative to coverage in a paired RNA-seq expression
library. This requires an expression library be constructed
and sequenced for the same strains/conditions used for 5
0
end mapping. The goal of this filter is to remove 5
0 ends
that represent degradation products present at relatively
low abundance compared to full-length transcripts, as well
as minor TSSs that result from spurious promoter-like
sequences (see Note 16).
For ease of execution, we suggest performing the first
two filters, then normalizing converted libraries relative to
nonconverted libraries, and finally performing the third
filter.
Use the script Collapsing_peaks_5_phos_OR_
mock.py to remove 5
0 ends that have read depths below
5 in all libraries and to retain only the 5
0 end with the
highest coverage in each 5 nt window.
2. Normalize converted libraries relative to nonconverted
libraries. This normalization is done empirically, based on the
assumption that the majority of the 5
0 ends present at this stage
are cleavage products that should be equally abundant in both
samples. When normalization is complete, we expect that the
total summed read depth of the converted libraries across the
genome will be greater than the total summed read depth of
the nonconverted libraries, since both libraries contain cleavage
sites and the converted library additionally contains TSSs. To
check if the expected relationship between the converted and
nonconverted libraries is present, calculate the log2 ratio of
relative abundance in converted relative to nonconverted
RNA 5
0
End Mapping
527
corrected for differences in sequencing depth relative to each
other, and all nonconverted libraries are corrected for differences in sequencing depth relative to each other. Differences in
depth between converted and nonconverted libraries are more
nuanced and will be addressed in the next step.
3.11 Filter to Obtain
Above-Background 5
0
Ends and Ensure
Correct Normalization
Between Converted
and NonConverted
Libraries
1. To identify 5
0 ends that are most likely to be biologically
meaningful, we suggest three filters (see Fig. 2):
(a) Remove 5
0 ends with normalized coverages of 5 reads
or less.
(b) Retain only the 5
0 end with the highest coverage in a 5-nt
window. This is useful because transcription initiation is
imprecise, and as a result, there will be substantial numbers of 5
0 ends mapping within a few nt both upstream and
downstream of the primary TSS (see Note 15).
(c) Retain only 5
0 ends that meet a read depth threshold
relative to coverage in a paired RNA-seq expression
library. This requires an expression library be constructed
and sequenced for the same strains/conditions used for 5
0
end mapping. The goal of this filter is to remove 5
0 ends
that represent degradation products present at relatively
low abundance compared to full-length transcripts, as well
as minor TSSs that result from spurious promoter-like
sequences (see Note 16).
For ease of execution, we suggest performing the first
two filters, then normalizing converted libraries relative to
nonconverted libraries, and finally performing the third
filter.
Use the script Collapsing_peaks_5_phos_OR_
mock.py to remove 5
0 ends that have read depths below
5 in all libraries and to retain only the 5
0 end with the
highest coverage in each 5 nt window.
2. Normalize converted libraries relative to nonconverted
libraries. This normalization is done empirically, based on the
assumption that the majority of the 5
0 ends present at this stage
are cleavage products that should be equally abundant in both
samples. When normalization is complete, we expect that the
total summed read depth of the converted libraries across the
genome will be greater than the total summed read depth of
the nonconverted libraries, since both libraries contain cleavage
sites and the converted library additionally contains TSSs. To
check if the expected relationship between the converted and
nonconverted libraries is present, calculate the log2 ratio of
relative abundance in converted relative to nonconverted
RNA 5
0
End Mapping
527