-
|
Dear DIANN community, At the moment I am searching multiple bacterial metaproteomes of an environmental time series. For each sample, I measured two biological replicates from the same day/system. Each sample was individually extracted, prepared and measured in a sequential manner. When changing parameters in DIANN 1.9.2 (linux command-line) I saw that some have quite an effect on protein and peptide recovery. Therefore, I hoped to ask if someone could help me to verify the settings for my purpose as I have a hard time to understand if my settings indeed fit the underlying samples and experiment. The main goal of this study is to scree the metaproteome in an explorative manner, trying to identify important and abundant proteins in the system. Before running the search, I generated a database like this: After that I used the generated database to search my samples: During trials I noticed quite some differences when adding: The main question I have here would be if I have to set --no-maxlfq whenever I use --no-norm. Also if it is indeed needed to put no-norm for my type of analysis. Also I was wondering if it is indeed valid to put relaxed-prot-inf in that case ? Moreover I'm not sure if I should use smart-profiling instead of rt-profiling ? Here the description when generating the database: and the description when analysing the samples with it: I want to thank you for any comment regarding that post. I have to admit I'm rather inexperienced in the field of metaproteomics which is why I invested some time into trying different settings and see what I get. THANK YOU |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 3 replies
-
|
Hi,
In general, a good idea is to keep settings default with changes as recommended in DIA-NN docs, and no other changes. In 99% cases this is the best approach. This means no M(Ox) or N-term(Ac) modifications.
Fine, except you don't want to include variable mods on the first try, as indicated above. If your data is dia-PASEF, makes sense to restrict the precursor charge range. In any case, makes sense to restrict the precursor m/z range to that of the experiment.
No.
I don't think so. You can always apply your custom normalisation in R on top of DIA-NN's normalised quantities.
Protein inference is in general quite 'unreliable'. So yes, I would keep in 'heuristical' but pay attention to the Protein.Ids column in your report and maybe rely on Proteotypic peptides only and respectively Genes.MaxLFQ.Unique - for this also a good idea to use --ids-to-names (replaces Gene names with sequence ids, better to avoid any confusion between orthoologues) for both predicted library generation and analysis.
No. |
Beta Was this translation helpful? Give feedback.
-
|
THANK YOU very much for your quick answer and your fantastic tool. Your comments are very helpful. Best, |
Beta Was this translation helpful? Give feedback.
-
|
Hello Vadim, Zooming in on this remark you made:
Why is that? It's the setting I've always used in other software (and in my uses of DIA-NN so far), and I often identify plenty of peptides with these modifications. Are your concerns with speed, quantitation, handling multiple modifications at once, or something else maybe? Kind regards, |
Beta Was this translation helpful? Give feedback.
-
If the other software does not ensure peptidoform confidence, it's unknown if the reported peptides ids are actually correct.
But in 99% of cases the number of proteins will stay the same, while the search takes longer. |
Beta Was this translation helpful? Give feedback.
Hi,
In general, a good idea is to keep settings default with changes as recommended in DIA-NN docs, and no other changes. In 99% cases this is the best approach. This means no M(Ox) or N-term(Ac) modifications.
Fine, except you don't want to include variable mods on the first try, as indicated above. If your data is dia-PASEF, makes sense to restrict the precursor charge range. In any case, makes sense to restrict the precursor m/z range to that of the experiment.
No.