Re: Mass++ Score & Hit Issues

[bbmis...@gmail.com]

Hi All,

I am using GC/MS Data (Thermo Raw Files to mzXML and so on...) to subject to Identification using Massbank.

Post-analyses, in the Results/ Report Summary, I receive by Email this Table:

No.	Query Name	Score	Hit	MassBank ID	Record Title	Formula	Exact Mass
1	Scan 0	0.5365	399	KZ000183	alpha-Methyl-DL-serine; GC-EI-TOF; MS; 3 TMS; BP:114	C4H9NO3	119.05824
2	Scan 1	0.5087	395	KZ000183	alpha-Methyl-DL-serine; GC-EI-TOF; MS; 3 TMS; BP:114	C4H9NO3	119.05824
3	Scan 2	0.5422	420	KZ000183	alpha-Methyl-DL-serine; GC-EI-TOF; MS; 3 TMS; BP:114	C4H9NO3	119.05824

and so on.

I have used samples that were MSTFA (and hence TMS groups)  + methoxyamine derivatized.

My Questions are as follows:

1. Which "matching scores" are acceptable ? Ofcourse 1.0 is not achievable with real data, right ? But for ALL of my matches I get score ranging from 0.2-0.6 only, not even 0.7 or 0.9, let alone 1. And what does this Score indicate ?

2. Secondly, how many hits are considered as a cut-off, usually ? 10 is sufficient to assign an identitiy/ annotation to a spectra ?

3. Depending on the answers to the above questions- what are the usual 2 cut-off values based on the Score and Hit to assign the 'real identity' with some level of confidence ?

Any advice and suggestions in this direction would be greatly appreciated !

Thanks and regards,
Biswa

[satstnka]

Hi,

 I'm Satoshi Tanaka, one of Mass++ developers at Shimadzu Corporation.
Thank you for using Mass++.

  MassBank is developed MassBank team at NAIST. (Nara Institute of Science and Technology.)
    http://www.massbank.jp/
 I recommend you asking questions about MassBank to get quick and correct answers.
You can get contact information from the following url:
    http://www.massbank.jp/en/contact.html

 However I asked MassBank team your question in this time.
Prof. Nishioka, a chief of MassBank team, said
Search score in MassBank is calculated using cosine correlation method.
It's standard method and used by NIST Mass Spectral Library too.
If query's peaks are detected as many as target spectrum's,
the degree of similarity is calculated by cosine correlation method.
 Methoxyamine (CH5ON), you used for testing, is quite a small molecule,
so the number of peaks must be few.
 In such a case, scores tend to be small.

 Though there is no standard value, you can say "it's similar" in general if
the score is more than 0.7.
 However it depends on cases.

 Prof. Nishioka recommended you to test compounds which is contained in MassBank
to find the threshold of similarity score.

regards,

--
Satoshi Tanaka