ParticipantID:        BertrandSamuel
Category:             category2
Authors:              Bertrand, Samuel(1) and Guitton, Yann (1,2,3,4,5)
Affiliations:         (1) Groupe Mer, Molécules, Santé-EA 2160, UFR des Sciences 
                          Pharmaceutiques et Biologiques, Université de Nantes, France 
                      (2) Université de Rennes 1, IRISA UMR 6074, Rennes, France
                      (3) CNRS, IRISA UMR 6074, Rennes, France
                      (4) Centre Rennes-Bretagne-Atlantique, Projet Dyliss, INRIA, 
                          Rennes Cedex, France
                      (5) Université de Nantes, LINA UMR 6241, Nantes, France

Automatic methods:    yes

Abstract
The challenge data were automatically treated using R, Rdisop, RMassBank as follow:
1- retrieving MS Data.
2- searching for NeutralLosses, adducts according to MS1 data. When no information 
were found ions were considered as [M]+, [M+H]+, [M+Na]+, [M+K]+, [M+H+solvent]+, 
[M+NH4]+ (if ammonia was added in the eluant), [M-H]-, [M+Fa]- (if Fa was added) and 
[M+AceticAcid]- (if AA was added); some manual interventions were done to simplify 
some complex detected ion relationship.
3- for each adduct possibility of each challenge molecular formula were deduced by 
Rdisop using CxHyO27N25P9F34Fe (modified according to neutral losses and adduct 
assigned) as maximum allowed atoms in the very strict limit of the MS accuracy. 
Each possible formula should pass the seven golden rules to be saved.
4- each possible formula were scored according to MS accuracy, Spectral Accuracy, MF
redundancy (multiply detected adducts from one MS spectra), type of adduct, the 
possibility to find Molecular formula related to MS2 ions from the MF itself.
5- only formula found in chemSpider/chebi/KnApSaCK were reported for challenges 1-42 
(no filtering according to existance were done for chellanges 43-48).
6- molecular formula were searched into various databases looking for CAS number, 
InChI, InChIKey, SMILES, Biological Sources: challenges 1-15 -> HMDB, ChEBI; 
challenges 16-31 -> DNP, DMNP, AntiBase, KNAPSACK, ChEBI; challenges 32-42 -> ChEBI, 
ChemSpider (No search was done for challenges 43-48).
7- for each compounds found in the data bases missing data were completed (as much as
possible) using OpenBabel, CTS, CACTUS, ChemSpider.
8- Scores were calculated according to Biological sources similarity between the 
challenge and the possible annotation using http://www.catalogueoflife.org (only for 
challenges 1-31)
9- MS2 similarity between simulated and measures MS2 were evaluated and scored using 
CFM-ID (due to time restriction many challenges of category3 were not completely 
calculated).
10- final scores was calculated according to MF score and possible annotation scores.

ParticipantID:        lridder
Category:             category2
Authors:              Ridder, Lars
Affiliations:         Wageningen University, Laboratory of Biochemistry,
                      Wageningen, The Netherlands
Automatic methods:    yes

Abstract
The challenge peak lists were converted to MAGMa input files, and processed
with MAGMa (http://www.emetabolomics.org/magma) using candidate molecules from
HMDB and PubChem (Ridder et al. 2012, Ridder et al. 2013)
MAGMa does not make use of searches in spectral libraries.
MAGMa was first run based on candidates from the Human Metabolites Database (HMDB).
When none of the HMDB candidates scored well, i.e. provide a satisfactory annotation 
of the MS/MS fragments, candidates from PubChem were added. In case of of very large 
numbers of PubChem candidates they were filtered based on "refscore" 
(Ridder et al. 2014). The elements Cl, Br and S were excluded from candidate formula 
where obvious from the isotope pattern in the MS spectrum. The score provided in the 
files submitted to CASMI represents the "refined ranking" in MAGMA as described in 
Ridder et al. (2013).

References:
Ridder, L.; van der Hooft, J. J. J.; Verhoeven, S.; de Vos, R. C. H.; van Schaik, R.;
Vervoort (2012) J. Rapid Commun. Mass Spectrom. 26, 2461-2471.
Ridder, L.; van der Hooft, J. J. J.; Verhoeven, S.; de Vos, R. C. H.; Bino, R. J.;
Vervoort, J. (2013) Anal. Chem. 85, 6033-6040.
Ridder, L.; van der Hooft, J. J. J.; Verhoeven, S. (2014) Mass Spectrom. 3, p. S0033

ParticipantID:        felicityallen
Category:             Category 1 and Category 2
Authors:              Felicity Allen, Russ Greiner, David Wishart
Affiliations:         Department of Computing Science University of Alberta, Canada
Automatic methods:    yes

Abstract
A list of candidate molecular formulae was obtained by querying PubChem
for all structures within the stated MS tolerances of the precursor 
mass (as obtained by combining the integer precursor mass stated in the
experimental details file with the MS1 file). In cases where multiple
precursor masses were listed, candidates and molecular formula were included
for all listed masses. All unique molecular formulae were collated from these 
structures.

The MS1 spectra were then predicted for each candidate molecular formula using
the emass program by A. Rockwood and P. Haimi [1].  These predicted spectra
were compared to the provided MS1 spectra (restricted to within 10 Da of the 
monoisotopic mass of the molecular formula), and an MS1_SCORE was produced 
for each molecular formula based on the closeness of this match. The scoring
metric used was:  
MS1_SCORE = ( (WP + WR + DP)_3ppm + (WP + WR + DP)_5ppm + (WP + WR + DP)_10ppm )/10
where 
WP = intensity weighted precision (0-100)
WR = intensity weighted recall (0-100)
DP = dot product (0-1) x 100

[1] Rockwood A. and Haimi P., "Efficient calculation of accurate masses of isotopic 
peaks.", Journal of the American Society for Mass Spectrometry, 17:3 p415-9 2006.	

A list of candidate structures was obtained by querying all of the following
databases for all candidates within the required mass ranges (determined as above):
HMDB  http://www.hmdb.ca/
ChEBI http://www.ebi.ac.uk/chebi/
ChEMBL https://www.ebi.ac.uk/chembl/
Metlin http://metlin.scripps.edu/
FOODB http://foodb.ca/
T3DB http://www.t3db.ca/
DrugBank http://www.drugbank.ca/
ECMDB http://www.ecmdb.ca/
YMDB http://www.ymdb.ca/
PubChem https://pubchem.ncbi.nlm.nih.gov
PlantDB Privately held list of 200,000 plant and plant-derived compounds.

For all candidate structures with an MS1_SCORE > 28, CFM-ID [2,3] was used
to predict an MS2 spectrum for all candidates at both a 10V and 20V energy level. 
The MS2_SCORE was computed for each to reflect the closeness of this match:

MS2_SCORE = (DP + 2JA + WP + WR)_10V + (DP + 2JA + WP + WR)_20V
where WP, WR and DP are as above, and
JA = Jaccard Score (0-1) x 100

[2] Allen F., Pon A., Wilson M., Greiner R., Wishart D., "CFM-ID: A web server for 
annotation, spectrum prediction and metabolite identification from tandem mass 
spectra", Nucleic Acids Research, Web Server Edition 2014.

[3] Allen F., Greiner R., Wishart D., "Competitive Fragmentatation Modeling of 
ESI-MS/MS spectra for putative metabolite identification", Metabolomics, 11:1, 
p98-110, 2015.

For all candidates, a DB_SCORE was produced according to which of the above databases 
it was found in, to make use of the provided information about metabolic origin. 
Consider the database names below as indicators, i.e. add the amount if in that 
database):

DB_SCORE (Challenge 1-15) = HMDB (+200) + METLIN (+100) + ChEBI (+20) + FooDB (+20) +
DrugBank (+20) + ECMDB (+10) + YMDB (+10) + ChEMBL (+10).
DB_SCORE (Challenge 6-31) = ChEBI (+100) + ECMDB (+100) + YMDB (+100) + PlantDB 
(+100) + Metlin (+50) + FooDB (+50) + ChEMBL (+10).
DB_SCORE (Challenge 32-42) = HMDB (+50) + ChEMBL (+100) + T3DB (+100) + METLIN (+100)
+ DrugBank(+100)
DB_SCORE (Challenge 43-48) = 0

An additional manual offset of +/-50 was also made in some cases to reorder the top 
couple of candidates. This was done where descriptions of compounds in the database 
provided additional information for or against a match e.g. where a particular 
species or biofluid was provided in the experimental data.

For Category 2, the results were ranked according to the sum of the above three 
scores: TOTAL_SCORE = MS2_SCORE + DB_SCORE + MS1_SCORE

For Category 1, the molecular formula was computed for each structure
from Category 2 and kept in the same order. The list was then processed
to remove duplicate entries, keeping only the highest ranked listing 
for each unique molecular formula.

ParticipantID:        avaniya
Category:	          Category 2
Authors:	          Vaniya, Arpana, Yan Ma, Dominique Ardura, Zijuan Lai, 
                      Mine Palazoglu, Sajjan Singh Mehta, Gert Wohlgemuth, 
                      Tobias Kind, Oliver Fiehn
Affiliations:	      Genome Center, University of California, Davis 
Automatic methods:    No

Abstract:	
The challenge peak lists were converted to MSP files and was used to search. The 
Human Metabolome Database (HMDB) and  perform MS/MS mass spectral library search 
against NIST 14, MassBank, METLIN, ReSpect. This search generated a list of candidate
structures which were used in Mass Frontier 7.0 to generate fragment ions. Fragment 
ions were assigned to the MS2 spectrum of a each challenge to find the structure 
which would assign the greastest number of peaks in the MS2 spectrum. These fragment 
ions were verified using MS2Analayzer, as a list of neutral losses were generated 
from the MS2 spectrum from each challenge. The reverse-dot product score from the 
mass spectral library search and the resulting Standard InChIs were reported for 
submission to CASMI. 

ParticipantID:        dverdegem
Category:             category2
Authors:		      Verdegem, Dries and Ghesquière, Bart
Affiliation:          Metabolomics Core Facility, Vesalius Research Center, 
                      VIB/KULeuven, Leuven, Belgium
Automatic method:     yes

Abstract
For all assignments, we used the MAGMa software [1] in combination with our own 
in-house metabolite database, which was compiled from several existing databases 
(Metlin, Kegg, HMDB, LMID and PubChem). MAGMA was run with optimized cost parameters:
Single bond break cost: 1.0
Double bond break cost: 2.7
Triple bond break cost: 3.0
Aromatic bond break cost: 2.7
Missing substructure penalty: 7.0
All other setting were kept as in the original version of the software.

[1] Ridder, Lars, et al. "Substructure-based annotation of high-resolution multistage
MSn spectral trees." Rapid Communications in Mass Spectrometry 26.20 (2012): 
2461-2471.

ParticipantID:        Fitch 
Category:             category1 and category2
Authors: 			  William Fitch (1)
Affiliations:         (1) Stanford University (bfitch@stanford.edu)
Automatic methods:    No 

Abstract
I used manual methods to generate all structures and formulas. I did not have a good 
formula generator that used isotopic ratios. I did use ChemCalc sometimes and 
Xcalibur’s spectrum simulator. MS/MS interpretation was purely manual with a few 
database searches for confirmation. The dereplication choices among isomers were 
highly subjective based on experience and likely abundance in stated samples.

MS data is not often adequate to select among isomers.  We need LC retention. You 
gave me that data but I didn’t know how to use it. I recently published a 
metabolomics paper (Rapid Comm Mass Spec 2013, 27, 2091) in which I hinted at methods 
for making interlab retention data reproducible using sets of internal standards and 
well identified components. Unfortunately I have not found funding to pursue this 
avenue further.

Two other ideas for assisting these structure determinations, especially the unknowns 
would be deuterium exchange and retention at 2 ph values. Determining the number of 
exchangeable hydrogens is simple to perform, can help select among dereplication 
isomers and gives extra clues with true unknowns. The retention at 2 pH experiment 
requires some standardization but has the potential to indicate acid base character 
of the unknown.