Supplementary Information

Heuristic Algorithm

The Heuristic Algorithm contains multiple steps listed below:

The heuristic is a conservative algorithm that attempts to select a method that is not over or under reporting protein IDs based on all four inference methods executed. Empirical thresholds are set for Parsimony and Peptide Centric (lower threshold) as well as for Inclusion and Exclusion (upper threshold) which helps guide the decision making process for selecting a recommended method. These thresholds are separate for Inclusion and Exclusion because thess two algorithms typically over report Protein IDs (Inclusion) or under report Protein IDs (Exclusion).

  1. First each of the four main inference methods is executed.
  2. A finite number (100) of false discovery rates are genereted within the range [0, fdr_threshold] by increments of fdr_threshold * 0.01. fdr_threshold is typically set to 0.05.
  3. We loop over each FDR and the following is done at each specified FDR:
  4. The mean of the number of proteins identified at each FDR is calculated across all four methods.
  5. The number of standard deviations each method is from the mean at each FDR is calculated.
  6. The distribution of the number of standard deviations from the mean is plotted for each method on the same graph. See Heuristic Plot for an example.
  7. The peak of each distribution (each inference method) is identified and the absolute value of the x-axis coordinate is taken of each value. These are the heuristic scores.
  8. A recommended method is selected using the following conditions
    • If the heuristic score for parsimony and peptide-centric is within 0.5 stdev of zero then both methods are recommended. If only one of the two methods is within 0.5 stdev of zero then that method is selected. If criteria is not met, continue to the next step.
    • If the heuristic score for inclusion and exclusion is within 1 stdev of zero then both methods are recommended. If only one of the two methods is within 1 stdev of zero then that method is selected. (This step is only applicable if step 1 criteria is not met).
    • If neither of these conditions above are met then the method with the heuristic score closest to zero (The least amount of stdev from the mean) is selected as the recommended method.

Inference Types

Inclusion

Inclusion simply maps all peptides to all possible proteins. In this model we allow peptides to map to multiple proteins. This approach is useful if you want to analyze all possible proteins (even those that do not have any distinguishing peptides).

img/inclusion.png

Parsimony

Parsimony is the process of taking the list of peptides and mapping them to the minimal set of protein identifiers available. This is a standard method that is good at limiting the overall number of proteins but still utilizing all selected peptides. This method allows shared peptides to map to more than one protein. Assuming that the proteins the peptides get mapped to also contain at least one unique peptide across the searches.

img/parsimony.png

Exclusion

Exclusion maps all peptides to all possible proteins but removes any peptide from the search that is not distinguishing. This means that if a peptide maps to more than one protein it will be removed. With this inference model the database selection is very important. Ideally the database selected for searches should have limited redundancy. The redundancy is computed on the database level, meaning if a peptide maps to more than one protein that is contained in the Fasta database then that peptide will be removed from the analysis entirely. Exception: If two or more proteins within the database map to the exact same set of digested peptides the algorithm will select the first listed protein and discard the others.

img/exclusion.png

Peptide Centric

For Peptide Centric inference all peptides are assigned to all possible proteins. Each peptide is then assigned a protein group based on the mentioned possible protein map. For protein group naming, the possible proteins for the peptides in the group are concatenated to a list separated by a semi-colon. This method is useful when there are a lot of shared peptides between multiple proteins. This will create new protein groups based on the shared peptides. This can sometimes more accurately represent the biological state.

img/peptide_centric.jpeg

First Protein

For the First Protein inference method each peptide gets assigned to one protein only. The protein that gets assigned to each peptide is the first listed protein. This is typically the first protein listed in the Fasta database file. This method is very simplistic but useful if you just want a quick way to get a sense of the protein FDR and are not worried as much about the peptide to protein mapping.

Protein Picker

Protein Picker is an algorithm that treats target and decoy proteins as pairs and is essentially target/decoy competition. If both the target and decoy proteins are identified from the searches when protein picker is run, then the target and decoy scores are compared with one another. The one with the better score is kept to continue on in the analysis while the one with the worse score gets filtered out of the analysis. This algorithm is integrated into other tools such as Percolator Protein Inference.

Protein Score Types

Score Type Description
Best Peptide Per Protein Uses the best scoring PSM as the overall score for a given protein. This can be beneficial to use when proteins of interest map to few peptides.
Multiplicative Log Multiplies all of the PSM scores together and then takes the log of the value (This only works for psm scores where lower is better).
Top Two Combined Takes the two best scoring peptides and applies Multiplicative Log to them to get the protein score.
Additive Adds all of the PSM scores together (This only works for scores where higher is better).
Down Weighted Multiplicative Log Multiplicative Log but normalizes by the number of PSMs per protein. This method attempts to normalize small and large proteins. In a typical multiplicative log a larger protein would have an advantage at having a higher score than a small protein.
Iterative Down Weighted Multiplicative Log Multiplicative Log but normalizes by the number of PSMs per protein in an iterative fashion. The more psm scores a protein has at each iteration each psm score will impact the overall protein score less and less. This method attempts to normalize small and large proteins. In a typical multiplicative log a larger protein would have an advantage at having a higher score than a small protein.
Geometric Mean Takes the geometric mean of the PSMs per protein.

Export Types

It is advised to use the Standard Export Types OR q_value

Standard Export Types

All standard export types report back protein leads if there is protein sub-setting from the specific tool. (Parsimony can provide protein subsets but in these export types we only choose to report the group lead). The tables below represent what the output of each export type will look like. Exports are all in CSV format.

  • peptide: This is a standard export type that reports back Protein, Score, Qvalue, and Peptide sequences in a square format. By default, peptide sequences are separated by a space " ".
Protein Score Q_Value Number_of_Peptides Identifier_Type GroupID Peptides
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 CGVEVTQTK EGLNVLQY#FISTHGAR FATSDLNDLYR IALASPDMIR IPQESGGTK LIPAGTGYAYHQDR MGAEAIQALLK NTLLHEQWCDLLEENSVDAVK RVDYSGR VADLFEAR VIDIWAAANDR VTAEDVLKPGTADILVPR
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 CQTCGYKFHEHCSTK FQMFQLIDIAR QTAQGMDYLHAK SASEPSLHR VFLPNKQR WHGDVAVKILK
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 GYLSPDLSK QTAQGMDYLHAK SASEPSLHR TFFSLAFCDFCLK
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 0.0 2 Reviewed 4 LYLLTQMPH YCWMSTGLYIPGR
HNRPU_HUMAN|Q00839 15.316094065486292 0.0 2 Reviewed 5 AEGGGGGGRPGAPAAGDGK LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
##TCAF2_HUMAN|##A6NFQ2 2.4079456086518722 0.3333333333333333 1 Reviewed 6 MEPTPVPFCGAK
  • psms: This is a standard export type that reports back Protein, Score, Qvalue, and PSMs in a square format. By default, PSMs are separated by a space " ".
Protein Score Q_Value Number_of_Peptides Identifier_Type GroupID Peptides
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 CGVEVTQTK EGLNVLQY#FISTHGAR FATSDLNDLYR IALASPDMIR IPQESGGTK LIPAGTGYAYHQDR MGAEAIQALLK NTLLHEQWCDLLEENSVDAVK RVDYSGR VADLFEAR VIDIWAAANDR VTAEDVLKPGTADILVPR
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 CQTCGYKFHEHCSTK FQMFQLIDIAR QTAQGMDYLHAK SASEPSLHR VFLPNKQR WHGDVAVKILK
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 GYLSPDLSK QTAQGMDYLHAK SASEPSLHR TFFSLAFCDFCLK
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 0.0 2 Reviewed 4 LYLLTQMPH YCWMSTGLYIPGR
HNRPU_HUMAN|Q00839 15.316094065486292 0.0 2 Reviewed 5 AEGGGGGGRPGAPAAGDGK LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
##TCAF2_HUMAN|##A6NFQ2 2.4079456086518722 0.3333333333333333 1 Reviewed 6 MEPTPVPFCGAK
  • psm_ids: This is a standard export type that reports back Protein, Score, Qvalue, and PSM Identifiers in a square format. By default, PSM IDs are separated by a space " ". Values in Peptides column will be the PSMid values from data input.
Protein Score Q_Value Number_of_Peptides Identifier_Type GroupID Peptides
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 13 14 15 16 17 18 19 20 21 22 23 24
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 1 2 3 4 5 8
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 3 4 6 7
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 0.0 2 Reviewed 4 10 9
HNRPU_HUMAN|Q00839 15.316094065486292 0.0 2 Reviewed 5 11 12
##TCAF2_HUMAN|##A6NFQ2 2.4079456086518722 0.3333333333333333 1 Reviewed 6 27
  • long: This is a standard export type that reports back Protein, Score, Qvalue, and PSM Identifiers in a long format. Each Peptide for each Protein is listed down the file This format is good for merging to exterior data to swap with the inference reference.
Protein Score Q_Value Number_of_Peptides Identifier_Type GroupID Peptides
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 CGVEVTQTK
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 MGAEAIQALLK
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 NTLLHEQWCDLLEENSVDAVK
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 RVDYSGR
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 FATSDLNDLYR
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 EGLNVLQY#FISTHGAR
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 LIPAGTGYAYHQDR
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 VADLFEAR
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 IPQESGGTK
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 IALASPDMIR
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 VTAEDVLKPGTADILVPR
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 VIDIWAAANDR
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 QTAQGMDYLHAK
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 VFLPNKQR
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 FQMFQLIDIAR
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 SASEPSLHR
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 CQTCGYKFHEHCSTK
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 WHGDVAVKILK
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 QTAQGMDYLHAK
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 GYLSPDLSK
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 TFFSLAFCDFCLK
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 SASEPSLHR
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 0.0 2 Reviewed 4 YCWMSTGLYIPGR
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 0.0 2 Reviewed 4 LYLLTQMPH
HNRPU_HUMAN|Q00839 15.316094065486292 0.0 2 Reviewed 5 LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
HNRPU_HUMAN|Q00839 15.316094065486292 0.0 2 Reviewed 5 AEGGGGGGRPGAPAAGDGK
##TCAF2_HUMAN|##A6NFQ2 2.4079456086518722 0.3333333333333333 1 Reviewed 6 MEPTPVPFCGAK

Legacy Export Types:

All Legacy export types report on the peptide level and can either report protein group leads, or it can report all proteins (Proteins that are subsets of another protein will be reported in these cases for Parsimony).

  • q_value: This is similar to the peptide export type other than that the peptide sequences reported will be put into a new column. This causes the resulting file to NOT be square.
Protein Score Q_Value Number_of_Peptides Identifier_Type GroupID Peptides
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 CGVEVTQTK EGLNVLQY#FISTHGAR FATSDLNDLYR IALASPDMIR IPQESGGTK LIPAGTGYAYHQDR MGAEAIQALLK NTLLHEQWCDLLEENSVDAVK RVDYSGR VADLFEAR VIDIWAAANDR VTAEDVLKPGTADILVPR
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 CQTCGYKFHEHCSTK FQMFQLIDIAR QTAQGMDYLHAK SASEPSLHR VFLPNKQR WHGDVAVKILK
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 GYLSPDLSK QTAQGMDYLHAK SASEPSLHR TFFSLAFCDFCLK
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 0.0 2 Reviewed 4 LYLLTQMPH YCWMSTGLYIPGR
HNRPU_HUMAN|Q00839 15.316094065486292 0.0 2 Reviewed 5 AEGGGGGGRPGAPAAGDGK LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
##TCAF2_HUMAN|##A6NFQ2 2.4079456086518722 0.3333333333333333 1 Reviewed 6 MEPTPVPFCGAK
  • q_value_all: This is similar to q_value except all proteins will be reported (Not just leads).
Protein Score Q_Value Number_of_Peptides Identifier_Type GroupID Peptides
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1 CGVEVTQTK EGLNVLQY#FISTHGAR FATSDLNDLYR IALASPDMIR IPQESGGTK LIPAGTGYAYHQDR MGAEAIQALLK NTLLHEQWCDLLEENSVDAVK RVDYSGR VADLFEAR VIDIWAAANDR VTAEDVLKPGTADILVPR
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 CQTCGYKFHEHCSTK FQMFQLIDIAR QTAQGMDYLHAK SASEPSLHR VFLPNKQR WHGDVAVKILK
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 2 GYLSPDLSK QTAQGMDYLHAK SASEPSLHR TFFSLAFCDFCLK
BRAF_HUMAN|P15056 35.24577101017814 0.0 3 Reviewed 2 GYLSPDLSK QTAQGMDYLHAK VFLPNKQR
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 GYLSPDLSK QTAQGMDYLHAK SASEPSLHR TFFSLAFCDFCLK
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 3 CQTCGYKFHEHCSTK FQMFQLIDIAR QTAQGMDYLHAK SASEPSLHR VFLPNKQR WHGDVAVKILK
BRAF_HUMAN|P15056 35.24577101017814 0.0 3 Reviewed 3 GYLSPDLSK QTAQGMDYLHAK VFLPNKQR
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 0.0 2 Reviewed 4 LYLLTQMPH YCWMSTGLYIPGR
HNRPU_HUMAN|Q00839 15.316094065486292 0.0 2 Reviewed 5 AEGGGGGGRPGAPAAGDGK LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
B3KX72_HUMAN|B3KX72 15.316094065486292 0.0 2 Unreviewed 5 AEGGGGGGRPGAPAAGDGK LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
Q96BA7_HUMAN|Q96BA7 6.907755278982137 0.0 1 Unreviewed 5 LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
##TCAF2_HUMAN|##A6NFQ2 2.4079456086518722 0.3333333333333333 1 Reviewed 6 MEPTPVPFCGAK
  • q_value_comma_sep: This is similar to q_value except proteins in the group of a lead will be listed in a separate column denoted "Other_Potential_Identifiers". Peptide identifiers are not shown.
Protein Score Q_Value Number_of_Peptides Identifier_Type GroupID Other_Potential_Identifiers
RPOC_SHIF8|Q0SY12 82.89306334778564 0.0 12 Reviewed 1
RAF1_HUMAN|P04049 70.7434325345954 0.0 6 Reviewed 2 ARAF_HUMAN|P10398 BRAF_HUMAN|P15056
ARAF_HUMAN|P10398 46.288402190472596 0.0 4 Reviewed 3 RAF1_HUMAN|P04049 BRAF_HUMAN|P15056
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 0.0 2 Reviewed 4
HNRPU_HUMAN|Q00839 15.316094065486292 0.0 2 Reviewed 5 B3KX72_HUMAN|B3KX72 Q96BA7_HUMAN|Q96BA7
##TCAF2_HUMAN|##A6NFQ2 2.4079456086518722 0.3333333333333333 1 Reviewed 6
  • leads: This is similar to q_value except Q values are not reported and only proteins passing the fdr threshold set in the parameters will be reported.
Protein Score Number_of_Peptides Identifier_Type GroupID Peptides
RPOC_SHIF8|Q0SY12 82.89306334778564 12 Reviewed {1} CGVEVTQTK EGLNVLQY#FISTHGAR FATSDLNDLYR IALASPDMIR IPQESGGTK LIPAGTGYAYHQDR MGAEAIQALLK NTLLHEQWCDLLEENSVDAVK RVDYSGR VADLFEAR VIDIWAAANDR VTAEDVLKPGTADILVPR
RAF1_HUMAN|P04049 70.7434325345954 6 Reviewed {2, 3} CQTCGYKFHEHCSTK FQMFQLIDIAR QTAQGMDYLHAK SASEPSLHR VFLPNKQR WHGDVAVKILK
ARAF_HUMAN|P10398 46.288402190472596 4 Reviewed {2, 3} GYLSPDLSK QTAQGMDYLHAK SASEPSLHR TFFSLAFCDFCLK
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 2 Reviewed {4} LYLLTQMPH YCWMSTGLYIPGR
HNRPU_HUMAN|Q00839 15.316094065486292 2 Reviewed {5} AEGGGGGGRPGAPAAGDGK LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
  • all: This is similar to q_value_all except Q values are not reported and only proteins passing the fdr threshold set in the parameters will be reported.
Protein Score Number_of_Peptides Identifier_Type GroupID Peptides
RPOC_SHIF8|Q0SY12 82.89306334778564 12 Reviewed {1} FATSDLNDLYR MGAEAIQALLK IPQESGGTK NTLLHEQWCDLLEENSVDAVK EGLNVLQY#FISTHGAR CGVEVTQTK LIPAGTGYAYHQDR VIDIWAAANDR VTAEDVLKPGTADILVPR IALASPDMIR RVDYSGR VADLFEAR
RAF1_HUMAN|P04049 70.7434325345954 6 Reviewed {2, 3} SASEPSLHR FQMFQLIDIAR WHGDVAVKILK QTAQGMDYLHAK CQTCGYKFHEHCSTK VFLPNKQR
ARAF_HUMAN|P10398 46.288402190472596 4 Reviewed {2, 3} SASEPSLHR QTAQGMDYLHAK GYLSPDLSK TFFSLAFCDFCLK
BRAF_HUMAN|P15056 35.24577101017814 3 Reviewed {2, 3} QTAQGMDYLHAK GYLSPDLSK VFLPNKQR
ARAF_HUMAN|P10398 46.288402190472596 4 Reviewed {2, 3} SASEPSLHR QTAQGMDYLHAK GYLSPDLSK TFFSLAFCDFCLK
RAF1_HUMAN|P04049 70.7434325345954 6 Reviewed {2, 3} SASEPSLHR FQMFQLIDIAR WHGDVAVKILK QTAQGMDYLHAK CQTCGYKFHEHCSTK VFLPNKQR
BRAF_HUMAN|P15056 35.24577101017814 3 Reviewed {2, 3} QTAQGMDYLHAK GYLSPDLSK VFLPNKQR
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 2 Reviewed {4} LYLLTQMPH YCWMSTGLYIPGR
HNRPU_HUMAN|Q00839 15.316094065486292 2 Reviewed {5} AEGGGGGGRPGAPAAGDGK LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
B3KX72_HUMAN|B3KX72 15.316094065486292 2 Unreviewed {5} AEGGGGGGRPGAPAAGDGK LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
Q96BA7_HUMAN|Q96BA7 6.907755278982137 1 Unreviewed {5} LQAALDDEEAGGRPAMEPGNGSLDLGGDSAGR
  • comma_sep: This is similar to q_value_comma_sep except Q values are not reported and only proteins passing the fdr threshold set in the parameters will be reported. Peptide identifiers are not shown.
Protein Score Number_of_Peptides Identifier_Type GroupID Other_Potential_Identifiers
RPOC_SHIF8|Q0SY12 82.89306334778564 12 Reviewed {1}
RAF1_HUMAN|P04049 70.7434325345954 6 Reviewed {2, 3} ARAF_HUMAN|P10398 BRAF_HUMAN|P15056
ARAF_HUMAN|P10398 46.288402190472596 4 Reviewed {2, 3} RAF1_HUMAN|P04049 BRAF_HUMAN|P15056
TCAF1_HUMAN|Q9Y4C2 19.048939464610452 2 Reviewed {4}
HNRPU_HUMAN|Q00839 15.316094065486292 2 Reviewed {5} B3KX72_HUMAN|B3KX72 Q96BA7_HUMAN|Q96BA7