--pernt-norm-factor-threshold for short RNAs and --dms option

[irinatu]

Dear All,
I have a question regarding a reasonable value for the --pernt-norm-factor-threshold parameter for relatively short RNAs. I used the --dms option, so the normalization is prepared separately for each nucleotide. What are reasonable threshold values for 120 nt or even shorter, such as 90 nt?
I would greatly appreciate any help.
Irina.

[Psirving]

Hi Irina,

Tagging @anthonymustoe and @lucaskearns, I'm not as familiar with the details of the new Shapemapper. Can one of you confirm that I'm explaining this correctly? This is a new limitation of ShapeMapper v2.2 with --dms. If this isn't already explained somewhere in the documentation outside of the --help message, it might be worth adding.

The normalization method assumes that the reactivities follow a certain distribution. This assumption is more valid with a larger sample size. That's why Shapemapper2 requires 20 data points per nucleotide by default. You can lower this value, but be aware that the normalization can have weird behavior, inflating or deflating a nucleotide relative to the others.

One way around this is to use a control RNA. From the same DMS reaction (so that they are probed identically), amplify a second larger RNA, preferably a well-structured single-state RNA. You can run a single Shapemapper2 job on multiple RNA targets, and the normalization factors are calculated once using all of the available nucleotides.

-Patrick

[irinatu]

Thank you for this information. Do you always add a well-known structured RNA to the experiment in addition to the target RNA? Is there any way to deal with it if we've already done the experiment for a target RNA alone. The coverage is high, but the RNA target is a bit longer than 100 nucleotides?

[Psirving]

No, we don't always add a control RNA. I usually only recommend this for very short RNAs, which is something we don't do often.

If the multiple small targets were probed together, you can combine them into 1 shapemapper job, and you shouldn't have to change the thresholds. If not, lowering --pernt-norm-factor will at least let you get a profile to think about, just keep in mind the caveats.

[anthonymustoe]

Hi Irina and Patrick, My apologies for the delayed response. Patrick, I agree with your response. Going below --pernt-norm-factor-threshold 10 does not make sense in my opinion. I think a value of 10 could potentially be ok, but it will be system and even replicate dependent. The reason why is because the normalization is going to be computed from essentially a single nucleotide at that point (the 9th most reactive out of 10), so if your modification rates vary experiment to experiment (they will) this could result in large variation in apparent normalization. Also, if your RNA has no single-stranded Cs (for example), then normalization of Cs will be based on the 9th most reactive C, which in this case will be base-paired and thus make it look erroneously like every single C is reactive. Thus, very important you think carefully about the data and whether it is sensible if using a lower nt threshold number. I also agree with Patrick's suggestion that in these cases it would be ideal to include a longer "reference" RNA. Tony

…

On Thu, Jan 30, 2025 at 8:12 AM Patrick Irving ***@***.***> wrote: No, we don't always add a control RNA. I usually only recommend this for very short RNAs, which is something we don't do often. If the multiple small targets were probed together, you can combine them into 1 shapemapper job, and you shouldn't have to change the thresholds. If not, lowering --pernt-norm-factor will at least let you get a profile to think about, just keep in mind the caveats. — Reply to this email directly, view it on GitHub <#61 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AK4NUWGHIYU7KZAE2RASZGD2NIXN7AVCNFSM6AAAAABVVBNBIOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMRUGYZDMNJRGE> . You are receiving this because you were mentioned.Message ID: ***@***.***>

-- Anthony Mustoe ***@***.*** (847) 530-2765