MetFrag Hands-on

Topic: Using MetFrag for compound identification with MS/MS data and additional information.

A online training presentation is available here.

---

In this hands-on session you will learn how to use MetFrag to annotate MS/MS spectra as a first step to identify a molecular structure given MS and MS/MS information. Furthermore, we will use additional experimental and meta data to support a putative identification.

---

MetFrag webservice workflow

In this example we have extracted a feature from a water (river) sample from a LC-MS/MS measurement with a precursor m/z 230.1162 at retention time 10.1 minutes. The data is acquired on a LTQ Orbitrap XL with a high mass accuracy (<5ppm) in positive ion mode. The adduct type of the selected precursor ion is known as [M+H]⁺.

Please download the prepared data:

• MS1: ms1_mz230.1162_rt10.1.txt
• MS2: ms2_mz230.1162_rt10.1.txt

---

Step 1 - Run initial MetFrag processing

1 a) Retrieve Candidates from database

• visit the MetFragWeb tool in your browser https://msbi.ipb-halle.de/MetFrag
• define database settings to retrieve candidates given the MS1 information:
  1. use the precursor m/z value and type to calculate the neutral monoisotopic mass
  2. check mass accuracy
  3. select PubChemLite in the "Local Databases" section as compound database
• start a candidate retrieval by clicking "Retrieve Candidates"

---

Go to live demo

---

• MetFrag searches candidates matching the information given by the "Database settings" (here: Neutral Mass and 5 ppm deviation)
• after the retrieval you can download the candidate list as CSV or XLS to get a first overview about the retrieved data set

---

1 b) Process candidates by performing in silico fragmentation and matching to MS/MS data

• use the "Fragmentation settings" tab to add the given MS2 peak list
• you can visualize the peak list by clicking on the "Show Spectrum" button
• keep the settings for the in silico fragmentation and start the processing by clicking "Process Candidates"

---

---

• MetFrag now generates fragments for each candidate up to the specified tree depth
• the fragments are mapped to the MS/MS peak list (based on mass) which is used to calculate a score for each candidate
• after the processing is finished you see the ranked candidates list in the "Results" tab
• here you have different possibilities:
  • you can filter candidates by explained peaks
  • investigate explained fragments and calculated scores for each candidate
  • download ranked candidate list as CSV or XLS file

---

---

Questions:

Q1: How many different molecular formulas are present?

Q2: What do you think is the correct molecular formula?

Q3: What else could you do to verify the molecular formula besides using the given MetFrag results?

--

Visit http://www.envipat.eawag.ch/index.php and verify your molecular formula.

---

Step 2 - Run MetFrag processing using molecular formula

2 a) Retrieve Candidates from database

• use the same settings as in 1 a) but add the molecular formula
• also select "Include references" when using PubChem

---

Go to live demo

---

2 b) Process candidates by performing in silico fragmentation and matching to MS/MS data

• use the same settings as in 1 b) and process the candidates

---

Questions:

Q4: Looking at the results, what has changed compared to using the monoisotopic mass as candidate filter?

Q5: Is the molecular formula helpful here?

---

Step 3 - Run MetFrag adding additional experimental information

3 a) Add the retention time data model to the MetFragWeb tool

• adding additional information available from the experimental context is often helpful to verify a putative identification
• we want to add retention time as additional experimental information

---

• MetFrag includes a retention time model
• linear correlation of n-octanol/water partition coefficient(logP) and retention time
• candidate logP is predicted by XLogP3(retrieved from PubChem) or calculated by CDK's XLogP
• rt_XlogP.csv contains a data set of measured rt and XLogP3 values of 254 Eawag standards:

---

• upload the data set to the MetFragWeb tool in the "Candidate Filter & Score Settings" tab using the "Retention Time" panel on the right side (direct download: rt_XlogP.csv)

---

• after the file upload set the retention time of the precursor and select XLogP3 as partition coefficient which is used for correlation
• this results in an additional scoring term in the scoring function of MetFrag

---

3 b) Process candidates by performing in silico fragmentation and matching to MS/MS data

• use the same settings as in 2 b) and process the candidates

---

Questions:

Q6: What has changed compared to the previous run?

Q7: Use the weight sliders in the "Results" tab. Does it change anything?

Q8: Is the retention time information helpful here?

---

Step 4 - Run MetFrag adding additional meta information

4 a) Add the additional scoring terms

• meta information can help to verify putative identifications depending on the experimental context
• however, you need to be careful when using this information which is not related to your acquired data
• in the "Candidate Filter & Score Settings" tab select the additional "Database Scoring Terms"
  • PubChemNumberPubMedReferences
  • PubChemNumberPatents

---

---

4 b) Process candidates by performing in silico fragmentation and matching to MS/MS data

• use the same settings as in 3 b) and process the candidates

---

Questions:

Q9: What has changed compared to the previous run?

Q10: Would the number of references and patents have helped for a metabolomics experiment?

Q11: Investigate the high intensity fragments of the first ranked candidate. Are they plausible compared to fragment structures of other candidates?

---

---

Step 5 - Search in spectral libraries

5 a) Investigate MS/MS peaks in MassBank

• visit MassBank EU (https://massbank.eu)
• select the "Peak Search" and add the most intense explained peaks

---

• hitting "Search" to find spectra with matching peaks in the database

---

Questions:

Q12: Investigate the results and compare them to your MetFrag result list. Any conclusions?

---

Step 6 - Combine Spectra library search and MetFrag

6 a) Enable spectral similarity in MetFrag

• in the "Candidate Filter & Score Settings" tab enable "Spectral Similarity"
• MetFrag will now query the MS/MS peak list against a spectral library mirror to search for similar spectra of known compounds

---

---

6 b) Process candidates by performing in silico fragmentation and matching to MS/MS data

• use the same settings as in 4 b) and process the candidates

---

Questions:

Q13: Discard the meta information scores to just use the results based on experimental data. Any conclusions?

---

Exercise

• visit the CASMI contest site (http://www.casmi-contest.org/2017/challenges_1-45.shtml)
• try to identify some of the compounds
• check your results here

---

Advanced exercise: using MetFrag on command line

• MetFrag can be used on command line to process batches of annotation tasks -- its called MetFragCLI
• parameter files for MetFragCLI can be created by web interface
• get your copy of MetFragCLI from
  http://ipb-halle.github.io/MetFrag/
• setup one example calculation to retrieve a set of valid parameter

---

---

Prepare one directory with the required files for each annotation task

~/course$ ls *
MetFrag2.4.2-CL.jar  MetFragWeb_Parameters.zip

data:
challenge-001-msms.txt  challenge-003-msms.txt  challenge-005-msms.txt  challenge-007-msms.txt  challenge-009-msms.txt
challenge-001-ms.txt    challenge-003-ms.txt    challenge-005-ms.txt    challenge-007-ms.txt    challenge-009-ms.txt
challenge-002-msms.txt  challenge-004-msms.txt  challenge-006-msms.txt  challenge-008-msms.txt
challenge-002-ms.txt    challenge-004-ms.txt    challenge-006-ms.txt    challenge-008-ms.txt

MetFragWeb_Parameters:
MetFragWeb_Parameters.cfg  MetFragWeb_Peaklist.txt  README.txt

---

Prepare one directory with the required files for each annotation task

• slightly adjust MetFragWeb_Parameters.cfg to use ionized precursor mass
• works well in conjunction with "PrecursorIonMode" option

# 1 for M+H and -1 for M-H
PrecursorIonMode = 1 
FragmentPeakMatchRelativeMassDeviation = 5.0
SampleName = MetFragWeb_Sample
MetFragCandidateWriter = XLS
DatabaseSearchRelativeMassDeviation = 5.0
FragmentPeakMatchAbsoluteMassDeviation = 0.001
MetFragDatabaseType = PubChem
ResultsPath = .
#NeutralPrecursorMass = 272.068624
IonizedPrecursorMass = 272.068624
MetFragScoreTypes = FragmenterScore
MetFragScoreWeights = 1.0
MetFragPreProcessingCandidateFilter = UnconnectedCompoundFilter,IsotopeFilter
IsPositiveIonMode = true
MaximumTreeDepth = 2
NumberThreads = 1
UseSmiles = true
PeakListPath = MetFragWeb_Peaklist.txt

---

Prepare one directory with the required files for each annotation task

• create the directories and populate with files

for x in `seq -f %03g 1 9`; do
 mkdir challenge-${x};
 cp data/challenge-${x}* challenge-${x};
 cp MetFragWeb_Parameters/MetFragWeb_Parameters.cfg challenge-${x};
 ln -s challenge-${x}-msms.txt challenge-${x}/MetFragWeb_Peaklist.txt;
done

---

Prepare one directory with the required files for each annotation task

• inject precursor mass

for x in `seq -f %03g 1 9`; do
 mass=`head -n1 challenge-${x}/challenge-${x}-ms.txt | cut -f1`;
 sed -i 's|IonizedPrecursorMass =.*|IonizedPrecursorMass ='${mass}'|g' challenge-${x}/MetFragWeb_Parameters.cfg
done

---

Run all MetFrag processes

for x in `seq -f %03g 1 9`; do
 cd challenge-$x;
 java -jar ../MetFrag2.4.2-CL.jar MetFragWeb_Parameters.cfg
 cd ..;
done

---