EP2460175A1

EP2460175A1 - Spectral analysis data processing

Info

Publication number: EP2460175A1
Application number: EP10739968A
Authority: EP
Inventors: Michael Paul Alfred May; Matthew James Kelly; Lkuo Konishi
Original assignee: Shimadzu Corp
Current assignee: Shimadzu Corp
Priority date: 2009-07-28
Filing date: 2010-07-26
Publication date: 2012-06-06
Anticipated expiration: 2030-07-26
Also published as: EP2460175B1; WO2011012845A1; GB0913122D0

Abstract

A method of determining an indication of the presence or absence of a substance in a sample. The method comprises receiving data generated from the sample, the received data comprising a plurality of data points, each data point having an associated value indicative of a mass or mass to charge ratio and obtaining theoretical data points associated with said substance. For each of the plurality of data points, a probabilistic score is determined, the probabilistic score being determined based upon a range of values, the range being defined relative to the associated value. The theoretical data points and the plurality of data points are processed to determine correspondence between the theoretical data points and the plurality of data points and a strength of the correspondence is determined, the strength of correspondence being based upon the probabilistic scores. The presence or absence of the substance is determined based upon the strength of the correspondence.

Description

SPECTRAL ANALYSIS DATA PROCESSING

The present invention relates to methods and apparatus suitable for analysing spectra. More specifically, but not exclusively, the present invention relates to methods and apparatus for analysing spectra obtained from a sample to determine one or more organic compounds present in that sample.

Chemical analysis of a sample is used in many industries where determination of the composition of a sample is important. For example, one such area where chemical analysis of a sample is used is in proteomics where the structure and function of proteins is studied.

One technique for chemical analysis of a sample is mass spectrometry. In mass spectrometry a sample is ionized to generate charged molecules or molecule fragments and the abundance of particular molecules or molecule fragments is determined according to the mass-to-charge ratio of the molecules or molecule fragments. The abundances of molecules or molecule fragments with particular mass-to-charge ratios can be processed to generate a spectrum and the mass-to- charge ratios indicated in the spectrum can be analysed to determine substances present in the sample.

For certain substances such as proteins it is possible to predict fragments that will be generated when the substance is fragmented based upon the method used to fragment the substance. It is possible to calculate masses of the predicted fragments and, based upon the calculated masses, predict ions that will be generated when the substance is analysed. The predicted theoretical ions can be compared to an experimental spectrum generated from a sample to determine a probabilistic measure that the particular substance is present in the sample.

However, often a sample will contain more than one substance or will contain the substance of interest together with contaminants. In such a case an experimental spectrum may not closely match a predicted spectrum. Additionally, fragmentation may not occur exactly as predicted for a particular substance and predicted spectral peaks may not occur or peaks may occur that are not predicted. Identification of a substance in a sample is therefore difficult.

A very large number of proteins are known. For any sample, the taxonomy of proteins in the sample may be known. For example, it may be known that all proteins in the sample originate from a particular organism. However a particular taxonomy may include a very large number of proteins. Identifying which proteins in a particular taxonomy are present in a sample is therefore computationally expensive. As such, it has proved difficult to provide fast methods for determining whether a particular protein is included in a sample, based upon a spectrum generated from the sample and a predicted spectrum of the protein of interest.

Tandem mass spectrometry machines are machines capable of performing multistage mass spectrometry in which a fragment identified in a first spectrum can be selected and further fragmented to produce a second spectrum. Where a large number of fragments are identified in a first spectrum, real-time analysis of all fragments by producing respective spectra may not be possible. Methods for directing the selection of fragments of a first spectrum which are to be used as a basis for the generation of respective second spectra are therefore desirable.

It is an object of some embodiments of the present invention to obviate or mitigate at least some of the problems set out above.

According to a first aspect of the invention there is provided a method of determining an indication of the presence or absence of a substance in a sample. The method comprises receiving data generated from the sample, the received data comprising a plurality of data points, each data point having an associated value. Theoretical data points associated with the substance are obtained. For each of the plurality of data points, a probabilistic score is determined, the probabilistic score being determined based upon a range of values, the range being defined relative to the associated value. The theoretical data points and the plurality of data points are processed to determine correspondence between the theoretical data points and the plurality of data points. A strength of the correspondence is determined, the strength of the correspondence being based upon the probabilistic scores. The presence or absence of the substance is determined based upon the strength of the correspondence.

In this way, data obtained from a sample and theoretical data associated with a substance is processed to determine the presence or absence of the substance in the sample. The probabilistic score determined for each data point is based upon a range of values determined with reference to a value associated with that data point. That is, the range of values is defined specifically with reference to the associated value. It has been found that using a range which is defined with reference to an associated value provides accurate indications of the presence or absence of a substance in a sample, and is preferred to a method, in which, for example, a range is predefined without reference to the value associated with a particular data point.

The value associated with each of the data points in the received data may be a mass or mass to charge ratio. Each data point may indicate an abundance of chemical species (e.g. ions or molecules) having the associated mass or mass to charge ratio. That is, the received data may define a spectrum such as a spectrum produced by a mass spectroscopy apparatus.

For each of the plurality of data points, the range may be a range centred upon the associated value. Where the associated value is a mass, a suitable range may be defined by the associated value plus or minus a mass value in the range 0.01 Da to 0.3Da. Alternatively the range may be defined as a proportion of the associated value.

Each of the theoretical data points may have an associated mass or mass to charge ratio, and each theoretical data point may indicate an abundance of chemical species having the associated mass or mass to charge ratio.

For each of the data points, the probabilistic score may be based upon a number of theoretical data points having an associated value within the range defined with reference to that data point. For each of the data points, the probabilistic score may be based upon the number of theoretical data points associated with the substance. Each theoretical data point may specify a probabilistic value for the associated value.

The theoretical data points may be associated with a class of substances of which the substance is a member. For example, where the substance is a protein, the class of substances may be a set of proteins occurring in organisms within a particular taxonomy or set of taxonomies.

Processing the theoretical data points and the plurality of data points to determine correspondence between the theoretical data points and the plurality of data points may comprise determining a shared peak count, the shared peak count indicating a number of the plurality of data points having an associated value that corresponds to an associated value associated with one of the theoretical data points. Correspondence between a data point and theoretical data points may be defined with reference to said range which is based upon the value associated with that data point.

Determining a strength of correspondence may comprise generating a further probabilistic score based upon the probabilistic scores associated with each of the data points and the shared peak count.

Generating the further probabilistic score may comprise multiplying the probabilistic scores associated with each of the data points. The method may further comprise enumerating a plurality of combinations and summing or multiplying the probabilistic scores based upon the enumerated combinations. Enumerating the plurality of combinations may comprise storing a first combination and performing one or more bit-shift operations to enumerate others of the combinations. Such bit shift operations can be implemented using software, or alternatively can be implemented in hardware, for example using a field programmable gate array (FPGA).

The further probabilistic score may indicate a likelihood that a number of theoretical data points indicated by the shared peak count have associated values which correspond to associated values of the received data points. The further probabilistic score may indicate a likelihood that a number of theoretical data points equal to or greater than the shared peak count by chance have associated values which correspond to associated values of the received data points.

The further probabilistic score may indicate a likelihood that a first plurality of the data points have an associated value that correspond to an associated value associated with one of the theoretical data points and others of the plurality of data points have associated values that do not correspond to an associated value associated with one of the theoretical data points.

For any set of data points generated from a sample, some of the data points will in general match theoretical data points associated with at least some known substances. The match between theoretical data points associated with a known substance and the data points generated from the sample may be caused by the presence of a known substance in the sample, or may be a chance match. That is, the data points generated from the sample are independent of the theoretical data points associated with the known substance unless the known substance is present in the sample, but matches will often be generated for known substances which are not present in the sample. Each of the further probabilistic scores is an indication of how likely it is that the observed degree of correspondence between the data points generated from the sample and the theoretical data points associated with a substance occurred by chance. If it is likely that the degree of correspondence occurred by chance then the confidence of the presence of the substance in the sample is reduced and if it is unlikely that the observed degree of correspondence occurred by chance then the confidence of the presence of the substance in the sample is increased.

The received data may be spectral data. Each data point in the received data may be a peak in the spectral data.

The sample may comprise at least one chemical entity selected from the group consisting of proteins, peptides, carbohydrates and metabolites. The substance may be selected from the group consisting of proteins, peptides, carbohydrates and metabolites.

The theoretical data associated with the substance may comprise spectral data. Each theoretical data point may be a peak in the spectral data.

According to a second aspect of the invention there is provided a method for processing spectral data, the processing comprising a plurality of processing phases. The method comprises executing processing of a first processing phase on a first processing element and executing processing of a second processing phase on a second processing element.

In this way, processing in said first and second phases can be carried out in parallel, thereby improving the efficiency of the computation.

The spectral data may be processed by the first processing phase at a first time and the spectral data may be processed by the second processing phase at a second time subsequent to the first time. The first and second processing elements may operate in series such that particular spectral data is first processed by the first processing element and subsequently processed by the second processing element. In this way particular processing to which the spectral data is to be subjected is divided between the first and second processing elements, thereby increasing its efficiency.

The first processing phase may process first spectral data and the second processing phase may process second spectral data, the second spectral data being different to the first spectral data. That is, where first and second different spectral data is to be processed, the first and second spectral data may be processed by different processing elements, thereby improving the efficiency of the processing.

The processing may comprise carrying out a method according to the first aspect of the invention. According to a third aspect of the invention there is provided a mass spectrometry data analyser for processing spectral data in a plurality of processing phases. The mass spectrometry data analyser comprises a first processor arranged to execute processing of a first processing phase and a second processor arranged to execute processing of a second processing phase.

According to a fourth aspect of the invention there is provided a method of processing a sample to determine an indication of the presence or absence of a substance in the sample, the substance having associated theoretical data. The method comprises receiving a first data value indicating a first degree of correspondence between the theoretical data and data generated from the sample at a first time. A second data value indicating a second degree of correspondence between the theoretical data and data generated from the sample is received at a second time. An indication of the presence or absence of the substance in the sample is determined based upon the first degree of correspondence and the second degree of correspondence.

In this way, the fourth aspect of the invention provides a method in which first and second data values are used in a process arranged to determine the presence or absence of a substance in a sample. The first and second data values are preferably data values of the same type and are determined from a sample in the same way.

Determining an indication of the presence or absence of the substance in the sample may comprise determining that the substance is present in the sample if both of the first data value and the second data value exceed a predetermined threshold. That is, the parameter indicated by the first and second data values may need to exceed the threshold for at least a predetermined time before it is determined that the substance is present in the sample.

Determining an indication of the presence or absence of the substance in the sample may be based upon a difference between the first and second data values. Determining an indication of the presence or absence of the substance in the sample may be based upon a rate of change indicated by the first and second data values.

The method may further comprise processing the first and second data values to determine whether to carry out processing to receive a third data value indicating a third degree of correspondence between at least one of the theoretical data points and data generated from the sample at a third time.

Determining whether to carry out processing to receive a third data value may be based upon rate of change between the first and second data values.

Determining whether to carry out processing to receive a third data value may be based upon a relationship between the first and second data values and a threshold value.

If it is determined not to carry out processing to receive a third data value, a further substance may be selected and the sample may be further processed to determine an indication of the presence or absence of the further substance in the sample.

The data generated from the sample may be spectral data. The data generated from the sample may comprise a plurality of peaks.

The sample may comprise at least one chemical entity selected from the group consisting of proteins, glycoproteins, peptides, carbohydrates and metabolites. The substance may be selected from the group consisting of proteins, glycoproteins, peptides, carbohydrates and metabolites.

The theoretical data associated with the substance may comprise spectral data. The spectral data may comprise a plurality of predicted masses.

The method may further comprise graphically displaying the first data value and the second data value to a user. The first data value and second data value may be displayed to the user whilst the sample is still being processed, or subsequent to the sample being processed. The data value may be generated at a plurality of times, and the data values at the plurality of times may be displayed as part of a graph indicating change in the data value over time.

According to a further aspect of the invention there is provided a method for analysing the composition of a sample. The method comprises generating a first spectrum based upon the sample and selecting a peak in the first spectrum to form the basis for further analysis. A second spectrum is generated based upon chemical entities represented by the selected peak. The second spectrum is processed to determine an indication of the presence or absence of a substance in the sample by carrying out a method according to the fourth aspect of the invention.

If the indication satisfies a first criterion the method may further comprise selecting a second peak in the first spectrum and generating a further second spectrum based upon chemical entities represented by the second peak. The further second spectrum may be processed to determine an indication of the presence or absence of a substance in the sample.

If the indication satisfies a second criterion, the method may further comprise generating a further second spectrum based upon chemical entities represented by the selected peak and processing the further second spectrum to determine an indication of the presence or absence of a substance in the sample.

If the indication satisfies a third criterion, the method may further comprise selecting a third peak in the second spectrum to form the basis for further analysis, generating a third spectrum based upon chemical entities represented by the third peak and processing the third spectrum to determine an indication of the presence or absence of a substance in the sample.

The method may further comprise graphically displaying the selected peaks to a user. For example, peaks may be displayed to a user and those peaks selected for further processing may be highlighted, for example using a predetermined colour to indicate selected peaks. If the indication satisfies a fourth criterion it may be determined that no further spectra based upon chemical entities represented by a selected peak should be generated.

Chemical entities may be, for example, fragments of a peak.

The terms likelihood and probabilistic score are used to indicate any probabilistic quantitative measure, including for example a probability, or a mathematical likelihood.

It will be appreciated that the various aspects of the invention can be implemented in any convenient way, including by way of suitable apparatus, and by way of computer programs. Such computer programs may be carried on appropriate computer readable media.

Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

Figure 1 is a schematic illustration of a system for spectral analysis according to an embodiment of the present invention;

Figure 2 is a graph showing an example spectrum generated by the system of Figure 1 ;

Figure 3 is a schematic illustration of a further system for spectral analysis according to an embodiment of the present invention;

Figure 4 is a series of graphs showing selection of ions from a spectrum and generation of spectra from the selected ions;

Figure 5 is a schematic illustration showing possible ion fragments of a peptide; Figures 6A and 6B are each a graph showing an example spectrum and amino acids corresponding to the distance between peaks of the spectrum;

Figure 7 is a flowchart showing processing to identify substances in a sample;

Figure 8 is a graph showing an example of variance in frequency of occurrence of ions at various masses;

Figure 9 is a flowchart showing processing to determine a likelihood that a particular substance is present in a sample;

Figure 10 is a schematic illustration of a method for determining a value used in the processing of Figure 9;

Figure 1OA is a schematic illustration of part of the processing of Figure 10;

Figure 11 is a schematic illustration showing an arrangement of part of the system of Figure 1 in further detail;

Figure 11A is a schematic illustration showing an alternative arrangement to that of Figure 11;

Figure 12 is a flowchart showing processing to determine the presence of a substance in a sample in a system using the arrangement of Figure 11 ;

Figures 13A to 13J are each a graph showing plots of a value determined according to the processing of Figure 12;

Figure 14 is a flowchart showing directed data acquisition to identify substances in a sample; and

Figures 15A and 15B are schematic illustrations of data processing arrangements suitable for use with the systems of Figure 1 and 3. Referring to Figure 1 , a detector 1 is arranged to detect ions generated from a sample 2. Ions are generated from the sample 2 by an ion source 3 which ionizes molecules in the sample. A mass analyser 4 is arranged to separate ions generated by the ion source 3 according to the mass-to-charge ratio (m/z) of the ions such that the detector 1 can measure a quantity of ions with a particular mass-to-charge ratio and provide data to a computer 5. The computer 5 is arranged to communicate with a database 6 to store data based upon the data received from the detector 1. A suitable mass analyser is the LCMS-IT-TOF mass spectrometer of Shimadzu Corporation, Kyoto, Japan, although any suitable mass spectrometry system may be used such as an LC-ESI (liquid chromatography - electrospray ionisation) ion trap TOF (time of flight) mass spectrometer.

Figure 2 shows an example spectrum generated by an apparatus such as the spectral analysis system of Figure 1. The spectrum has a plurality of peaks such as peaks 10, 11. Each peak is defined by a mass-to-charge ratio indicated on the x-axis and an associated intensity indicated on the y-axis. A peak with a relatively high associated intensity such as peak 10 indicates a relatively high abundance of ions with the mass-to-charge ratio associated with the peak 10, while the peak 11 indicates a relatively low abundance of ions having the mass-to-charge ratio associated with the peak 11.

A spectrum such as the spectrum of Figure 2 can be used to identify a substance contained within the sample 2 of Figure 1. The sample is first processed to cleave the substance along some of its bonds to generate fragments. When the substance is analysed using the system of Figure 1 , each fragment generated by the cleavage generates a peak in a spectrum such as peaks 10 and 11 of Figure 2. It is possible to predict fragments that will be generated from cleavage of a known substance based upon the method used to cleave the substance. The corresponding mass-to-charge ratio of the predicted fragments can also be predicted. The peaks in a spectrum generated from a sample can be used to identify a substance present in the sample based upon predicted fragments for a plurality of substances. In some embodiments where the sample is a protein, the sample is cleaved using a restriction enzyme (for example Trypsin) which breaks the sample at predictable points along the amino acid chain. The analysis of a spectrum, such as that shown in Figure 2, where the spectrum is obtained from analysis of a protein is known as peptide mass fingerprinting (PMF).

From the preceding description it will be appreciated that it is possible to predict a spectrum that a particular substance will generate when analysed by a system such as the system of Figure 1. However, in practice, spectra generated from a sample may be generated from a sample containing more than one substance, or from a sample containing contaminants. Fragmentation may not occur exactly as predicted and with electrical and chemical noise, large numbers of peaks may be generated. Identification of substances present in a sample may therefore be difficult.

A tandem mass spectrometry system for identification of substances is shown in Figure 3. The system of Figure 3 can be viewed as an extension of the system of Figure 1 as components 1A to 5A of Figure 3 correspond to components 1 to 5 of Figure 1 and are arranged in such a way that the system of Figure 5 can operate in the same way as the system of Figure 1. In the system of Figure 5, an ion selector 17 (which in some embodiments takes the form of an ion trap such as a quadrupole ion trap) is arranged to receive ions from mass analyser 4A. Computer 5A communicates with ion selector 17 to direct the selection of an ion, separated from other ions by mass analyser 4A, for further fragmentation by an ion fragmentation component 18. Ions generated by the ion fragmentation component 18 are further separated by the mass analyser 4A and detected by detector 1 A in the same way as described above with reference to the system of Figure 1. The selection of ions by ion selector 17 after separation by mass analyser 4A and subsequent analysis may be repeated a plurality of times.

An ion selected by ion selector 17 shall be referred to as a precursor ion, and ions generated by ion fragmentation of the selected precursor ion (carried out by the ion fragmentation component 18) shall be referred to as product ions. It can be noted that an ion may be both a product ion of a precursor ion and a precursor ion of several product ions. The process of generating a spectrum using a system such as the system of Figure 1 shall be referred to as MS¹. The process of generating spectra in general shall be referred to as MS" where n - 1 indicates the number of precursor ions that are in the chain of ions fragmented to generate the current spectrum in the system of Figure 3. For example MS² shall be used to refer to generation of one or more spectra from one or more precursor ions identified in an MS¹ spectrum (i.e. the ion selector 17 operates on ions of a single spectrum generated by the mass analyser 4A).

Figure 4 illustrates ion selection and spectra generation according to the system of Figure 3. Two MSⁿ⁺¹ spectra 20a, 20b are generated from respective precursor ions 21a, 21b indicated in an MSⁿ spectrum 22. Spectrum 20a is generated from an ion 21a and spectrum 20b is generated from an ion 21 b. Spectra 20a, 20b can be used to identify particular ions present in the sample used to generate MS" spectrum 22. For example, analysis of the spectrum 20a allows information to be obtained as to the ion responsible for the peak 21a in the spectrum 22. Such information is useful in determining the composition of the sample which resulted in generation of the spectrum 22.

Figure 4A shows the computer 5 of Figure 1 in further detail, although it will be appreciated that the computer 5A of Figure 3 generally has the same form. It can be seen that the computer comprises a CPU 5a which is configured to read and execute instructions stored in a volatile memory 5b which takes the form of a random access memory. The volatile memory 5b stores instructions for execution by the CPU 5a and data used by those instructions. For example, in use, output from the detector 1 may be stored in the volatile memory 5b.

The Computer 5 further comprises non-volatile storage in the form of a hard disc drive 5c. The output from the detector 1 may be stored on the hard disc drive 5c. The computer 5 further comprises an I/O interface 5d to which are connected peripheral devices used in connection with the computer 5. More particularly, a display 5e is configured so as to display output from the computer 5. The display 5e may, for example, display a representation of the output from the detector 1. Additionally, the display 5e may display graphical representations of results of processing the output from the detector 1. Input devices are also connected to the I/O interface 5d. Such input devices include a keyboard 5f and a mouse 5g which allow user interaction with the computer 5. A network interface 5h allows the computer 5 to be connected to an appropriate computer network so as to receive and transmit data from and to other computing devices. The CPU 5a, volatile memory 5b, hard disc drive 5c, I/O interface 5d, and network interface 5h, are connected together by a bus 5i.

A study to look at the number of different proteins that are identified by a single peptide in the human taxonomy has found that 43% of peptides uniquely define a single protein and 94% of peptides identify less than 6 proteins. Identification of a particular peptide ion generated from a protein, using MS² processing according to the system of Figure 3, and as shown in Figure 4, can therefore be useful for identifying proteins in an original sample as often one or two peptides can uniquely identify a protein.

Where the sample 2 is a protein, the cleavage process will generate a plurality of peptides. These peptides are then further fragmented by the ion fragmentor 18 which may be a collision cell. Figure 5 shows fragments that may be generated from a peptide 14. Lines 15 indicate bonds of the peptide 14 that may be broken and the arrows of lines 15 indicate corresponding ions that are generated after the peptide is processed by the ion source 3 of Figure 1. One method of breaking bonds in a sample which may be implemented by the ion fragmentor 18 is low energy Collision Induced Dissociation (CID) and it is known that using CID, peptides and polypeptides tend to fragment into b and y ions. That is, using CID₁ bonds such as peptide bond 16 between a CO and an NH group tend to be broken and ions b₂ and y₂ are generated. Peaks indicating ions with mass-to-charge ratios corresponding to ions b₂ and y₂ can therefore be predicted when substance 14 is fragmented using CID and the predicted peaks can be used to identify substance 14.

Figures 6A and 6B show spectra that are predicted to be generated from a peptide with amino acid sequence FIAGGER comprising amino acids represented by their standard IUPAC abbreviations F, I₁ A, G, E and R. The predicted spectra are based on a fragmentation of the substance using CID and as such b and y ions are predicted to be generated and detected. Each peak in each of the spectra corresponds to a predicted ion fragment, with Figure 6A showing predicted b-ions and Figure 6B showing predicted y-ions. The mass of an ion b₃ corresponds to the mass of an ion b₂ with an additional mass corresponding to the mass of amino acid I. Ion b₃ can therefore be deduced to have the same structure as ion b₂ with an additional I amino acid group. Other additional amino acid groups can also be determined by the difference in mass between detected ions and are indicated between ion peaks. Corresponding y-ions for each b-ion are shown in Figure 6B.

MSⁿ processing, for n greater than or equal to 1, for identification of a substance will now be described at a high level with reference to Figure 7. At step S1 a sample is received and at step S2 ions are generated from the sample. The received sample may be a polypeptide or mix of polypeptides, or other molecules such as metabolites. In the case where n is greater than 1 the received sample is a precursor ion generated from MS^{n 1} processing, which is further fragmented to generate product ions, and the substance to be identified is the precursor ion.

At step S3 the ions are separated according to the mass-to-charge ratio of the ions and at step S4 the separated ions are detected and the abundance of each ion, generated at step S2, is determined. The determined abundances of ions, together with their respective mass-to-charge ratios produce a spectrum such as the spectrum described above with reference to Figure 2. At step S5 the generated spectrum is processed to determine at least one substance present in the sample received at step Sl

In general terms, and as is described in further detail below, to determine if a particular ion or polypeptide D is present in a sample S, based upon a set of peaks Ps associated with a spectrum generated by MS" processing of the sample, a probability based score providing an indication of a probability that the peaks Ps match predicted peaks associated with any substance of the same type as D purely at random at least as well as the peaks Ps match a set of predicted peaks P_D associated with D is generated. A substance can be considered to be of the same type as D based upon any suitable criterion. For example, if it is in a taxonomy or set of taxonomies associated with D (such as the same class, the same family or the same species) or has a particular number of peptides or ion fragments based upon the number of peptides or ion fragments associated with D or both.

In further detail, the set of predicted peaks P_D is determined based upon the expected peaks associated with D, the expected peaks being determined from the mass-to-charge ratio of expected ions or peptides associated with D determined as described above with reference to Figure 3. A shared peak count k indicating the number of peaks in the set P_s that match a peak in the set P₀ is determined. For each number of peaks h less than k, a probability R(h) of any h peaks in Ps matching predicted peaks associated with any substance of the same type as D is generated, based upon a probability associated with each peak in the set Ps occurring in any spectrum generated from a sample which includes a substance of the same type as D. The likelihood R(h) for each number of peaks h is summed to determine a likelihood L of any substance of the same type as D matching the peaks in the set Ps less well than the match between P_s and P₀- A value P_vaι(k), determined by the value (1 - L), indicates the likelihood that peaks associated with any substance of the same type as D will match the set of peaks P_s at least as well as the set of peaks P₀ match the set of peaks Ps (i.e. will match k or more peaks). The value P_vaι(k) is a measure of the quality of the match between D and the spectrum generated from the sample. Smaller values of P_va,(k) indicate a better match (and thus a greater likelihood that D is in the sample) than larger values of P_vaι(k).

The likelihood associated with each peak in the set P_s occurring in any spectrum generated from a sample is determined from a probability density function (PDF) associated with the type of D. A PDF value for a peak having a particular mass-to- charge ratio is generated from a database of known substances of the same type as D by determining predicted fragments for each known substance in the database and determining how likely it is that a spectral peak will occur at the mass-to-charge ratio of the peak generated from the sample, based upon the mass-to-charge ratio of the predicted fragments. Each mass-to-charge ratio that may be detected by the particular instrument used will have a particular PDF value, based upon the number of fragments predicted from substances in the database having the particular mass- to-charge ratio. High accuracy machines are able to detect a large number of separate mass-to-charge ratios, and the variation between PDF values associated with different mass-to-charge ratios is likely to be greater than for low accuracy machines. A PDF table associated with various types (that is taxonomies or sets of taxonomies or of particular sizes of substance) may be generated and stored and used to look up PDF values for particular peaks. For MS1 peaks PDF values are specific to a particular restriction enzyme such as trypsin. Additionally, PDF values may be determined based upon a predetermined number of cleavages that may be missed. Larger numbers of possible missed cleavages generate larger numbers of possible peptides. One possible missed cleavage has been found to be a suitable number.

Whilst it is possible to generate a PDF table for each possible number of fragments, in practice a summed PDF table is produced for some ranges of numbers of fragments. For low numbers of peptides per protein, PDF values vary considerably and as such, individual numbers of peptides each have their own PDF table. Suitable ranges of numbers of peptide fragments in a polypeptide per PDF table have been found to be one PDF table per number of peptide fragments for numbers of peptides in the range 1 to 19, two PDF tables for numbers of peptides in the range 20 to 30 and increasing ranges of numbers of peptides for peptide numbers greater than 30. In total 53 PDF tables have been found to be suitable for ranges of numbers of peptides although other numbers of PDF tables and ranges of numbers of peptide in a polypeptide per PDF table may be suitable. PDF tables for ion fragments are determined in a similar manner, based upon the frequencies of different numbers of ion fragments.

Predicted and experimentally observed peaks tend to fall in clusters in a small range of masses out of the total range of possible masses. An example mass cluster is shown in Figure 8 for the range 998.3 to 998.8 with the number of predicted peaks indicated on the y-axis and the mass indicated on the x-axis. A continuous line 25 indicates the actual number of predicted peaks for each mass and a line 26 indicates the kernel density estimation of the PDF. It can be seen that the actual number of predicted peaks for a particular mass within the cluster varies considerably, with a point 27 indicating that peaks indicating a mass of approximately 998.46 are relatively common compared to a point 28. Some spikes are caused by exact isobar peptides, both resulting from peptides with the same amino acid constituent parts and resulting from peptides with the same elemental composition. It can be seen that the kernel density estimation of the PDF fails to properly take into account the variance shown by the line 25.

Processing to determine a likelihood that a particular substance D is present in the sample S based upon the set of peaks P_s determined from a spectrum generated by MS" processing of the sample, for n greater than or equal to 1, will now be described in detail with reference to Figure 9. At step S10 the set of peaks P_s determined from the sample S is received, each peak p, in the set P_s having an associated mass-to- charge ratio m, (for i in the range 1 to the total number of peaks p in the set Ps). In some embodiments all generated ions have a charge of 'V and as such m, indicates mass. This may be for example due to the ions having been pre-processed to form a singly-charged peak list. The set of peaks P₅ may include all peaks present in the spectrum or may include a subset of the peaks present in the spectrum selected, for example, a subset of peaks of greatest intensity. At step S11 the set of predicted peaks P_D associated with the substance D is received and at step S12 a tolerance δ is received. The tolerance δ is a tolerance used in the determination of matches between peaks in the set Ps and peaks in the set P_D, and more particularly is a tolerance between masses associated with peaks in the set Ps and masses associated with peaks in the set P₀ required for peaks in the sets Ps and P₀ to be considered to match. At step S13 the number of peaks k in the set P_s that match a peak in the set Pp within the tolerance δ is determined.

At step S 14 a peak pi in the set Ps is selected and at step S15 a value C(p,) is determined for the peak p, according to equation (1 ): C(_Pi)

Where:

N_mass is the PDF value associated with D at a particular mass indicated by "mass"; and

MaxMass is a value indicating the maximum mass for a PDF value in the PDF table associated with D.

The denominator of the fraction of equation (1) gives the sum of all PDF values in the PDF table associated with D and the numerator gives the sum of the values in the PDF table with a mass within tolerance δ of the mass m, associated with the peak p,. The value C(p,) therefore is the probability of a peak with mass m,, within tolerance δ, occurring in a spectrum generated from any substance of the type associated with D.

At step S16 a value C_≥1(p,) is determined for the peak p, according to equation (2):

C₂₁(P₁H -(I -C(P,))' ₍₂₎

Where:

C(Pi) is the value determined at step S15 according to equation (1); and y is the number of fragments associated with D and therefore the number of peaks in the set P₀.

The value (1-C(p,)) is the probability that any single peak associated with a substance of type D does not match with the peak p,. The value (1-C(p,))^y is therefore the probability that y peaks associated with a substance D do not match a peak with mass m,, derived by assuming that matches are entirely random and as such will follow the statistics of Bernoulli trials. It follows that the value C_≥1(p,) is the probability that one or more peaks of the y peaks associated with a substance of type D matches the peak p,.

At step S17 a check is carried out to determine if there are more peaks in the set P_s that have not been processed according to the processing of steps S15 and S 16. If it is determined that there are more peaks to be processed then processing continues at step S14 where a previously unprocessed peak is selected from the set P₅. Otherwise it is determined that a probability of a peak matching at least one peak in a substance of type D has been determined for each peak in the set P_s and processing continues at step S18.

At step S18 a counter variable h is initialised to 0 and at step S19 a check is carried out to determine if the value h is equal to the value k. If the value h is not equal to the value k then at step S20 a value R(h) is determined according to equation (3): of h matches π isH ^c*w^><π jeP(ⁱ-^c»W) O) where:

C≥i(p,) is the value determined at step S16 according to equation (2);

i, j are in the set {1 p};

H is the set of h peaks that are considered to be a match in the particular combination; and

P is the set of peaks not in H such that P U H = {1 , ... ,p} and P D H ={ }.

A value C_≥1(p,) associated with each of the p peaks in the set P_s is therefore used in each of the calculations inside the summation. For each different combination of h matches, the set H (and correspondingly the set P) used within a particular iteration of the summation is different

The value given by the term inside the summation of equation (3) is the probability of a particular combination of h peaks in the set Ps matching predicted peaks associated with any particular substance of the type associated with D, together with the remaining (n - h) peaks in the set P_s not matching peaks associated with the particular substance of the type associated with D. The summation of equation (3) sums across all possible combinations of h matches. The value R(h) is therefore a probability of any combination of h peaks in the set P₅ matching peaks of a substance of the type associated with D, and the remaining peaks in the set Ps not matching a peak of the same substance.

At step S21 the counter variable h is incremented and processing returns to step S19. The loop of steps S19 to S21 generates a value R(h) for each combination of h peaks where h is less than k. That is, the loop of steps S19 to S21 generates a probability of peaks in the set P₅ matching peaks of a substance of the type associated with D less well than the set of peaks Ps matches the set of peaks P₀. If it is determined at step S19 that h is equal to k then at step S22 a value P_va,(k) is determined according to equation (4): where R(h) is the value determined according to equation (3).

The summation of equation (4) gives a likelihood of a match between peaks of any substance of the type associated with D and peaks in the set P_s that is not as good as the match between peaks in the set P₅ and peaks in the set P₀. That is, the summation provides a likelihood of a match of any h peaks, where h is less than k. The value P_vaι(k) is therefore the likelihood of a match between peaks in the set P_s and predicted peaks of a substance of the type associated with D that is at least as good as the match between the peaks in the set P_s and the peaks in the set PQ. The value Pvai(k) is a measure of the quality of the matches between the set P_s and the set P_D. A greater value of P_vaι(k) indicates a greater likelihood that the peaks in the set Ps match peaks of any substance of the type associated with D. As such, where the value of P_vaι(k) is high, there can be less confidence that matching an arbitrary k peaks in the set Ps to peaks associated with the substance D is a reliable indicator of the strength of the match between the sets of peaks P₀ and P_Sl and can be used as an indicator of whether or not the substance D is present in the sample S.

It has been described above that the value R(h) of equation (3) is determined for values of h less than k in order to determine the value P_vaι(k). It will however be appreciated that a value R(k) can be determined according to equation (3) where k is the peak match count between D and the set P_s. The value R(k) is itself a reliable indicator of the strength of the match between the sets of peaks P₀ and Ps and can be used in the place of or as well as the value P_vaι(k) in any of the methods described.

A value Score(k) can be determined from the value P_vaι(k) according to equation (5) below.

Score(k) = -/o£(P_val(k)) (5)

Greater values of Score(k) indicate a greater likelihood that D is present in the original sample (given the definition of P_vaι(k) as set out above), and the value Score(k) can be thresholded to determine the presence or absence of D in the sample. The threshold may be a predetermined value or may be determined as the minimum value that allows a substance to be identified with a particular confidence based upon the selected parameters (including for example the taxonomy). It will of course be appreciated that the value P_vaι(k) may be thresholded and used in the determination in the same way as Score(k).

A value Score(k) may be determined for each of a set of peptides or polypeptides Δ and the associated value Score(k) for each member of Δ used to determine the elements of Δ that are most likely to be present in the sample S. The set Δ may be determined by first performing a peptide mass fingerprint (PMF) technique with the set of peaks Ps and sets of predicted peaks associated with database proteins. Any database protein with a predetermined number of predicted peaks that match peaks in P₅ within a predetermined threshold (that is that have a shared peak count greater than the threshold) is added to the set Δ and therefore considered for processing to determine a value Score(k) where k is equal to the number of matching peaks.

It can be noted that the value P_vaι(k) does not depend upon the particular k peaks that are matched between P_s and P₀ and as such for any substance D' that is of the same type as D that has any combination of k peaks matching peaks P_s, the value P_vai(k) will be the same as the value P_vaι(k) for the substance D. A value R_patte_m(k) can be calculated to indicate a likelihood of the particular pattern of k matches between the set of peaks P_s and the set of peaks P₀ associated with a substance D according to equation (6):

where the right hand side of equation (6) is the same as the term inside the summation of equation (3), but here the set H includes the k peaks of the set Ps which matched peaks associated with the substance D.

The value R_pattem(k) takes into account likelihoods of the actual matching peaks. The value of equation (6) can be used to resolve which substance of two substances D and D¹, both of the same type, is the most likely to be present in a sample in the case where D and D' both have the same number of matching predicted peaks to the set Ps, depending upon the values R_patte_m(k) generated based upon each of D and D'.

It will be appreciated that the calculation of R(h) of equation (3) is computationally expensive. Indeed in general, for large values of p and relatively large values of k (where p is the number of peaks in the set P_s), the number of possible combinations of k matches will be very large, given the normal combinations formula Cf . Each possible combination requires p multiplications to calculate R(h). For a typical value of p of 60 and k of 12, the total number of computations is 8.4x10¹³. As will be discussed in further detail below, the value Score(k) associated with a substance D can be used to direct real-time data acquisition and ion selection for further processing such as the ion selection illustrated in Figure 6. However, determination of the value R(h) of equation (3) and therefore the value Score(k) of equation (5) in around 10 milliseconds as is required for real-time data acquisition in a standard way would require computation of the order of 10⁵ Teraflops and is therefore impractical with current computing power. A fast method for determining the value R(h) (and therefore the value P_vaι(k)) will therefore now be described.

In general terms, the set Ps is divided into q subsets T₁ to T_q of approximately equal size, where q is approximately equal to the square root of the number p of peaks in Ps. The size of each subset is referred to in the following description as N₁.

For each subset i, an associated value R(t) is determined, for all possible values of t in the range 0 to N₁ (or 0 to k where k is less than N₁). This corresponds to all possible combinations of k or less matched peaks in the subset.

A particular value inside the summation of equation (3) can be determined by selecting a single value R(t,) associated with each of the subsets T₁ to T_q where i is equal to 1 to q and indicates the associated subset, such that the sum of the values t, is equal to the number of matches k between predicted peaks associated with D and peaks in the set Ps- The value R(h) can therefore be determined by evaluating the summation of equation (3), each iteration of the summation being evaluated with reference to appropriate values of R(t,).

The method for determining the value R(h) described above is now illustrated by way of an example with reference to Figure 10. The example of Figure 10 is based upon a value of k (indicating the number of matching peaks) equal to 5 and a value of p (indicating the number of peaks in the set Ps) equal to 15. The set P_s is shown in Figure 10, where its elements are indicated by 15 boxes 30, 5 of which are indicated as '1 ' and 10 of which are indicated as ¹O' as determined by the values k and (p-k) respectively. Each of the boxes 30 represents a peak in the set P_s as indicated by labels 31 and the '0' or '1 ' indication in each box indicates whether the corresponding peak is considered to be a matching peak or a not matching peak. The particular combination of '0' and ¹V indications of combination 32 corresponds to one of the possible combinations of k matches required for the summation of equation (3).

As described above, the set of peaks are divided into approximately V15 disjoint sets (3 sets in this case) of approximately equal size T₁, T₂ and T₃ where subset Ti corresponds to peaks p₁₅ to P₁₁ , subset T₂ corresponds to peaks p₁₀ to p₆ and subset T₃ corresponds to peaks p₅ to P₁ (i.e. T₁ U T₂ U T₃ = P_s; T₁ H T₂= {}; T₁ D T₃ = {}; T₂ D T₃ = {})•

Processing is carried out to determine sets of values for each of the sets T₁ to T₃, corresponding to a value R(t) for each possible number of matching peaks in the set. For example a value T₁₁ shown in Figure 10 is determined by summing values ^₁₁, tii₂)--- . tu_n each corresponding to a possible combination of a single matching peak in the set T₁, and each corresponding to one of the possible combinations inside the summation of equation (3). Each value t₁₁₂ t_11n is determined by multiplying values

C_≥Xp,) for each p, indicated as "T (i.e. that is considered to be a match) with values (1 - C_≥1(p,)) for each p, indicated as '0' (i.e. that is considered to not be a match). For example, the value t₁₂₂ in set T₁₂ is determined by the calculation shown in (7): t,₂₂ = C_≥1 (p₁₅ )^χ (l - C_≥1 (p_M ))x C_≥1 (p₁₃ )^χ (l - C_≥1 (p₁₂ ))^χ (l - C_≥1 (p₁₁ )) (7)

All values T₁, for i equal to 1 to 3 and j equal to 0 to 5 are determined. Each of the possible values of equation (3) is given by a value determined by multiplying a value Ti_a, T₂b and T_3c such that (a+b+c) = 5. For example the value given by (Tn x T₂₂ x T₃₂) gives one of the possible values R(h). Predetermining the values in the sets of values T₁₁ saves a large number of multiplication operations as the multiplication operations are not repeated each time a value t_Uf is used to determine one of the possible combinations.

All possible bit combinations in each of the sets T₁, can be determined using a bitshift operation. Bitshift operations included in languages such as C and C++ are fast and the speed of the operation can be exploited to quickly enumerate all possible combinations in a set T₁,. Use of a bitshift operation to enumerate all possible combinations of the set Tn of Figure 10 is shown in Figure 1OA.

An array of values Pi corresponding to the value C_≥1(p,) for each peak is determined and an array of values P₀ corresponding to the value 1 -C_≥1(p,) for each peak is determined. The set T₁₁ corresponds to all combinations of matches and non- matches in the set of peaks p₁₅ to P₁₁ in which a single peak is matched and all other peaks are not matched. An initial arrangement of values satisfying the condition required for the set is enumerated, shown as U ₁₅. The value ti₁₅ is determined by multiplying the corresponding value from the set P₁ for each bit of t₁₁₅ that has a '1' indication and multiplying the corresponding value from the set P₀ for each bit of U ₁₅ that has a '0' indication, and the value tn₅ is stored. The next arrangement of values. t₁₁₄, that satisfies the condition of the set T₁₁ is generated from t₁₁₅ by shifting the set of values one place to the left, using the bitshift operation (indicated by « in Figure 10A) and the value ^ ₁₄ is determined by multiplying values from the sets P₁ and P₀ selected in the same way as for the enumeration t₁₁₅. All other enumerations are determined in the same way. The value T₁₁ is determined by multiplying each of the enumerations t₁₁₅ to tm. It will be appreciated that by appropriately initialising an arrangement of values for the set I_n and performing, in turn, as many bit shifts as there are elements in the set tn all necessary arrangements of values can be effectively enumerated.

The techniques described above for calculating values required in the calculation of R(h) can be efficiently implemented in software given that they provide a convenient mechanism for enumerating required combinations which relies upon the bit shift operator which is provided in the instruction set of most microprocessors and can be executed in a low number of clock cycles. In alternative embodiments, the various combinations may be enumerated using an appropriate logic circuit which can be arranged to generate the necessary combinations and use logical operators to determine the multiplications that are to be carried out to determine the value R(h). Such a logic circuit can be conveniently implemented using a Field Programmable Gate Array (FPGA). The fast method for determining values R(h) and associated values Score(k) and P_vaι(k) described above with reference to Figure 10 allows the value Score(k) to be used to direct real-time data acquisition and ion selection for further processing. Systems and methods in which the value Score(k) or other scoring metrics such as R(k) can be used will now be described.

Referring to Figure 1 1 a particular arrangement of the sample 2 and ion source 3 of Figure 1 is shown (it will be appreciated that the arrangement of Figure 11 can be similarly applied to the sample 2A and ion source 3A of Figure 3). The sample 2 is first passed through a liquid chromatography column 35 where molecules in the sample are separated according to their retention time through the chromatography column 35. An electrospray 36 corresponding to ion source 3, 3A generates ions from the molecules that have passed through the chromatography column 35. The generated ions are passed to the mass analyser 4 of Figure 1 (or correspondingly the mass analyser 4A of Figure 3) for processing as described above. The electrospray 36 generates ions constantly as molecules of the sample pass through the chromatography column 35 and the mass analyser 4 therefore receives ions generated from molecules having substantially the same retention time.

A further arrangement of the sample 2 and ion source 3 of Figure 1 is shown in Figure 11A. The system of Figure 11A is a Matrix-assisted laser desorption/ionization system (MALDI). The sample 2 and LC column 35 of Figure 11A correspond to the sample and LC column of the system of Figure 11. In the system of Figure 11 A, molecules separated according to their retention time through the chromatography column 35 are spotted onto a plate 36A in which different spots correspond to different retention times. The retention time of each spot can be determined based upon the location of the spot on the plate. The plate may be stored and processed at a later time (sometimes referred to as offline processing).

Processing to determine peptides or polypeptides present in a sample in a system including the arrangement of either of Figures 1 1 and 11 A will now be described with reference to Figure 12. In general terms, a value Score(k) is determined a plurality of times for a substance of interest D based upon a set of peaks P₅ generated from the sample at each time. Each of the determinations of the value Score(k) shall be referred to as a scan. The set of peaks P_s may be a set of peaks generated independently of the previous set of peaks or may be a cumulative set of peaks determined based upon the set of peaks P_s from a previous scan or a plurality of previous scans, together with data determined in the current scan. In the case where Ps is a cumulative set of peaks the signal to noise ratio will generally be improved.

In the system of Figure 11 , as time passes molecules with different chemical properties may be analysed at different times due to the different retention times of molecules and it may be desirable to reset the data periodically, for example when a determination about the presence of a particular substance in the sample has been made. The methods described are equally applicable to processing with either of the systems of Figure 11 and Figure 11 A.

The value Score(k) (or any other suitable metric including R(k)) is used to determine whether D is present in S in the usual way, by determining whether the value Score(k) associated with D exceeds a threshold. Additionally, values of Score(k) from preceding scans can be used to make decisions in relation to the substance D.

In more detail, at step S25 of Figure 12 data associated with the substance D is received for processing. As previously described, the substance D is a known peptide or polypeptide and has an associated set of predicted peaks Po- At step S26 the set of peaks P_s determined from the sample S is received and at step S27 the value Score(k) is determined for the substance D as described above with reference to Figure 9. At step S28 it is determined if further processing is required based upon an updated set of peaks Ps' determined from the sample at a subsequent scan after the set of peaks P_s was determined. The set of peaks Ps¹ may be either a cumulative set of peaks based on Ps or a fresh set of peaks as described above. Conditions for determining whether further processing is required are described below with reference to Figures 13A to 13J. If it is determined at step S28 that further processing is required then at step S29 the set of peaks P₅' is determined and processing continues at step S27 where a further value Score(k) is determined based upon the updated set of peaks P_s'. Otherwise at step S30 no further scans based upon substance D are performed. A determination of the presence or absence of substance D in the sample S may be output at step S30 or it may be determined that no determination can be made. Alternatively, a further substance may be selected based upon the result of the scans for substance D₁ for example a MSⁿ⁺¹ ion which is useful in identifying substance D.

It will be appreciated that any suitable metric calculated from the data other than Score(k), where the metric preferably uses more than one peak, can be used. Examples of suitable metrics include P_vaι(k) and R(k) described above, or known metrics such as a MASCOT score or a de novo sequencing score.

Referring now to Figures 13A to 13J, ten graphs showing plots of the value Score(k) are shown. In each of Figures 13A to 13J, a line 37 indicates the value of the threshold for determining the presence of a substance in a sample (i.e. the threshold which Score(k) is required to exceed). The threshold may be set based upon a significance test. A suitable threshold is a value that determines an event is significant if it would be expected to occur at random with a frequency of less than 5%. The value of the threshold shown in Figures 13A to 13J is 4.5 indicating a 1 in 31623 chance of the observed quality of match having occurred by chance.

Each graph is determined based upon a set of peaks P_s determined from a sample S and based upon a substance D and associated set of predicted peaks P₀- Each plot indicates a value Score(k) determined at intervals of 0.25 seconds indicating the interval time between scans. It should be noted that the scale of the time axis in each of Figures 13A to 13F varies between the Figures 13A to 13F.

Referring first to Figures 13A to 13D, each of the graphs of Figures 13A to 13D indicate that the value Score(k) has exceeded the threshold 37 and subsequent scans show that the value Score(k) remains greater than the threshold. It can therefore be determined that each of the substances for which the value Score(k) is shown in each of Figures 13A to 13D do not require further scans to be performed and that there is a reasonable degree of certainty that the substance D is present in the analysed sample. Indeed, a determination for each of the substances of Figures 13A to 13D can be made at a point 38A, 38B, 38C and 38D respectively where it can be seen that the value Score(k) has exceeded the threshold and the value Score(k) has remained substantially constant for 2 scans In some embodiments the number of scans required during which the score remains constant before a determination is made may be greater than 2, for example the number of scans required may vary according to the value Score(k). Additionally, in some embodiments further scans may be desirable so that the value Score(k) attains a higher score indicating a greater degree of confidence in the indication.

The graphs of Figures 13E and 13F indicate that the value Score(k) has not exceeded the threshold 37. The value Score(k) in Figure 13E has a value of approximately 3 at a time 1sec and the value of Score(k) remains constant for a further 1.25secs and 5 scans. Since the value Score(k) does not change for 5 scans it is unlikely that the value Score(k) will increase further and it can be determined that no further scans should be performed. The value Score(k) in Figure 13F decreases with a sharp gradient to a value 0 at a point 39 and again it can be determined that no further scans should be performed. Since the value Score(k) did not exceed the threshold 37 no positive determination can be made about the substance D for which the Score(k) is indicated in each of Figures 13E and 13F.

The graphs of Figures 13G and 13H each indicate that the value Score(k) has not exceeded the threshold 37 after 2secs and the value Score(k) is no longer increasing. The graph of Figure 131 shows the value Score(k) reaching a peak after 2.5secs before decreasing after 3.5secs. It may therefore be determined that it is no longer desirable to perform further scans for each of the substances D of Figures 13G to l.

The graph of Figure 13J indicates that the value Score(k) exceeds the threshold 37 at a point 40. However, the value Score(k) decreases below the threshold 37 at a point 41 and the determination of exceeding the threshold may therefore be considered to be less confident that if the value Score(k) remained above the threshold 37 for a predetermined number of scans subsequent to exceeding the threshold. Each of the substances D of Figures 13E to 13J may be used for directed data acquisition to select further substances which may confirm the presence of a substance related to substance D, such as a protein of which D is a peptide ion, as will be described in further detail below with reference to Figure 14.

Various properties of the value Score(k) across a plurality of repeat scans at different time intervals can be used in the determination of whether further scans are required or desirable. For example, the rate of change of the value Score(k) (i.e. the gradient of the plot of the value Score(k)) across a plurality of scans may be used, or the direction of change of the value of Score(k) may be used. The rate of change may be used in combination with the value of Score(k), or any other useful measure of the value Score(k) may be used.

In a system such as that described above, the set of peptides or polypeptides Δ may be determined based upon the retention time of the molecule at the particular time. A predicted value for the retention time of each peptide or polypeptide in the database can be stored and used to determine peptides or polypeptides that are expected to be passed from the chromatography column 35 to the electrospray 36 of Figure 11 at a particular time.

Use of the value Score(k) to direct real-time data acquisition using the MS" system of Figure 5 will now be described with reference to Figure 14. In general terms, a value Score(k) is determined for substances in a set Δ based upon a spectrum generated from a MS^{n 1} precursor ion (i.e. based upon an MSⁿ spectrum). If the value Score(k) exceeds a predetermined threshold for at least one substance D in Δ then a list of substances that may have been a precursor ion of the substance D in the MS"^'1 spectrum is generated. Based upon the list of substances that may have been a precursor ion of D, a list of further precursor ions is identified that either would be useful in determining the substance that resulted in the identified precursor ion or that should not be MS" processed. Referring to Figure 14, the processing is divided into a part A which is non-directed data acquisition processing and a part B which is directed data acquisition processing.

At step S35 a precursor ion ψ is selected from the MS"^'1 spectrum. The precursor ion may be selected based upon, for example, intensity of a peak associated with the precursor ion. At step S36 an MS" spectrum is generated from ψ and at step S37 a value Score(k) is generated for each substance in a set Δ in the usual way as described above with reference to Figures 7 and 9. At step S38 a check is performed to determine whether the value Score(k) associated with at least one substance D in Δ exceeds a threshold of 4.5. Each substance D identified at step S38 is a possible identification of the precursor ion ψ. If it is determined that the value Score(k) associated with each substance D in Δ is less than the threshold then processing continues to step S35 where a further precursor ion ψ in the MS"^'1 data is selected. Otherwise it is determined that useful information has been obtained from the MS" data associated with ψ and data directed acquisition can commence.

At step S39 a set Precursor_D (which is a set of proteins in the case where n is equal to 2) is determined based upon those substances D in Δ whose value Score(k) exceeded the threshold and at step S40 a set Product_D (which is a set of peptides in the case where n is equal to 2) is determined based upon the set Precursor_D. The set Precursor_D is a set of substances that are possibly identified by the presence of a substance D in the MS"^"1 spectrum (i.e. substances that may have been present in the sample from which the MS"^'1 spectrum was generated - it should be noted that the set Precursor_D does not necessarily correspond to a set of substances selected for MSⁿ⁺¹ analysis) and the set Product_D is a set of MS"^"1 ions whose identification is useful in confirming the presence of a substance in the set Precursor_D and are therefore a shortlist of masses for preferential MS" analysis. As described above 94% of peptides define less than 6 proteins, so in the case where n is equal to 2 where the set Precursor_D is a set of possible proteins and each D in Δ is a peptide, each D whose value Score(k) exceeds the threshold will in general identify less than 6 possible proteins. The set Product_D may be determined by identifying product ions that are predicted from substances in the set Precursor_D. For example in the case where n is equal to 2 and the set Precursor_D is a set of proteins for which a peptide D is predicted, the set Producto is a set of peptides other than D which are predicted from the predicted cleavage of proteins in the set of proteins Precursor_D. The set Product_D may be determined by identifying predicted peptides that uniquely identify a protein in the set Precursor_D or in any other suitable way.

At step S41 an ion ψ' is selected from the MS^M spectrum by ion selector 17 of the system of Figure 5 based upon the set Product_D. The ion ψ' may be selected by determining a mass-to-charge ratio of an ion in the set Product_D and selecting an MS"^"1 ion that has a corresponding mass-to-charge ratio within a predetermined threshold. At step S42 an MS" spectrum is generated from the ion ψ' (which is itself a precursor ion given that it is used to generate an MS" spectrum) and at step S43 a value Score(k) is generated for each substance in a set Δ'. The set Δ' may be selected based upon the set Product_D or may be determined in any other suitable way. The set Δ' indicates substances that are likely to provide useful information relating to the composition of a sample. At step S44 a check is carried out to determine whether the value Score(k) associated with at least one substance D' in Δ' exceeds a threshold of 4.5. If it is determined that there is no substance D' in the set Δ' that has an associated value Score(k) that exceeds 4.5 then processing returns to step S41 where a further precursor ion is selected. Otherwise at step S45 the set Precursor_D and the set Producto are updated based upon the substances D' identified as having a value Score(k) that exceeds the threshold.

The processing described above with reference to Figure 14 is particularly useful in the case where n is equal to 2 (i.e. MS¹ and MS² spectra are generated). In such a case, it is often possible to identify a protein in a sample based upon a single peptide identification (i.e. a single product ion identification). In such a case it may be desirable to confirm the presence of the protein by confirming the presence of other expected peptides. In other cases identification of a single peptide may identify a small number of proteins that may be present in the sample (in 94% of cases, less than 6). Identification of a further peptide generated from the sample (and identified by a peak in the MS¹ spectrum) will often identify which of the possible proteins is in fact present in the sample and caused the first identified peptide to be present in the MS¹ spectrum.

It has been described above that the set Precursor_D is a set of substances that are possibly identified by the presence of a substance D in the MS"^'1 spectrum and the set Product_D is a set of MS"^'1 ions whose identification is useful in confirming the presence of a substance in the set Precursor_D. In an alternative embodiment the set Product_D may be a set of MS"^'1 ions whose identification is not useful. In such a case, the selection of a precursor ion ψ' of step S41 based upon the set Producto may comprise selecting an ion whose mass-to-charge ratio does not correspond to the mass-to-charge ratio of an ion in the set Product_D. Such processing may be used, for example, where the set Precursor_D is a set of proteins that have already been identified, and the set Productp may be, for example, a set of peptides that are unique to an already identified protein (i.e. a member of the set Precursor_D). In such a case, identification of peptides in the set Product_D would not aid further identification of substances present in the input sample. Alternatively, both a set of MS"^"1 ions that are useful and a set of MS"^"1 ions that are not useful may be stored (together with corresponding sets of precursor substances) and used in the selection of MS^{n 1} precursor ions.

Graphical displays may be used to indicate results from or progress of MS processing, for example during MS processing or after MS processing or during or after MSⁿ processing, where n > 1. For example, graphs such as the graphs of Figures 13A to 13J may be generated and displayed to a user. A graphical display (for example displayed on a display of the computer 5 of Figure 1) may be arranged to indicate candidate peptides and/or ion series of candidate peptides in a tabular format. For example, ions selected to form the basis of further processing using MSⁿ⁺¹ processing may be indicated using particular colours and peptides may be coloured dependent upon the strength of confidence that the peptide is present in the sample. In this way, results of processing to identify substances in a sample can be displayed in a convenient way which allows a user to quickly and easily determine important results of the processing. Additionally, a list of peptides that have been identified as present in the sample may be displayed to a user, which may be updated during an MS run, and a list of proteins that may be present in the sample based upon the identified peptides may similarly be displayed to a user. Confidence that the protein is present in the sample may similarly be indicated using different colours.

Similarly, a graphical display may be used to indicate decisions relating to selection of ions for MSⁿ⁺¹ processing made by the system. For example, a hierarchy of ions which have been selected for MSⁿ⁺¹ processing from an MSⁿ spectrum may be displayed to a user so that the user is able to quickly determine a path of ion selections which result in a current MSⁿ⁺¹ ion selection. The tree may again indicate confidence of peptide identification in the tree using, for example, colours.

The systems described above with reference to Figures 1 and 5 have been described with reference to a single flow of control carrying out the described processing. Two alternate embodiments which use parallel processing (using multiple computers or processors) are shown in Figures 15A and Figure 15B.

In the embodiment of Figure 15A₁ three processors 45, 46, 47 arranged to process different data sets in parallel are shown. A first processor 45 is arranged to receive and process a first set of data from the detector 1 at a first time according to the methods set out above. The processor 45 corresponds to the computer 5 of Figure 1 or computer 5A of Figure 3. If at a second time a second set of data is available for processing (for example a further peak or set of peaks eluting from an LC column) and the processor 45 has not completed processing to identify substances in the sample based upon the first data set, the second processor 46 processes the second data set in the manner set out above. A further processor 47 is provided to process further data sets when both processors 45, 46 are processing previously received data sets. Further processors can be provided as required.

In the embodiment of Figure 15B, three processors 50, 51 , 52 are arranged to process data sets in serial. A first processor 50 is arranged to process a first data set received at a first time. The first computer 50 is arranged to perform a first part of the processing to identify substances present in a sample, such as pre-processing spectral data and/or the selection of peaks described above. Examples of processing that may be carried out by the first computer include charge deconvolution, de- isotoping and peak picking. When the first processor has completed the first part of the processing that the first processor 50 is arranged to perform, the first processor outputs the result of the processing to a second processor 51 which is arranged to perform a second part of the processing, such as processing selected peaks to identify the set of substances Δ in the database to be further processed (for example based upon a shared peak count, as described above). Once the first processor has output to the second processor based upon the first data set, the first processor is arranged to perform the first part of the processing for a second data set. A third processor is arranged to receive output from the second processor and to perform a further part of the processing, such as generating a value Score(k) for each substance in the set Δ. Multiple processors can be arranged in this way to each perform a part of the processing described above. In the arrangement of Figure 15B each processor can be arranged to perform a particular function, and each processor can be optimised for the particular function. For example the first processor 50 may be arranged to receive the signal and interpret the signal, The interpreted signal may be output to the second processor 51 which is arranged to determine a set of substances based upon shared peak count. The set out substances may be output to the third processor 52 which is arranged to generate a score P_vaι(k) for each substance received from the second processors 51.

It will be appreciated that the arrangement of Figure 15A can be combined with the arrangement of Figure 15B to further increase the processing of multiple data sets simultaneously.

Results showing the performance of the method of generating a value Score(k) will now be described.

Table 1 shows the results from a computation on an equal 25OfMoI binary mixture of GLaD derivatised yeast Hexokinase (KIBYHA) and human α-lactalbumin (LAHU). The column "Protein ID" shows proteins identified as significant hits, the column "Score" shows the value Score(k) determined according to the processing set out above, the column "nPeptides" shows the number of peptides in the identified protein, the column "MASCOT score" shows the score generated using the MASCOT MIM of Matrix Science Inc. on the same data multiplied by ten and the column "Rank" shows the rank of the proteins based upon Score(k).

TABLE 1

In Table 1 a number of almost identical proteins have the same score. The almost identical proteins are grouped together in Table 1. It can be seen that the order of substances based upon the value Score(k) matches the order of substances as determined by the MASCOT score closely. It can be seen that KIBYHA and LAHU are both correctly identified as present in the sample by MASCOT and by the value Score(k) having rank 1 and rank 3 respectively. ALF-YEAST is a genuine contaminant of the sample and is correctly identified by both systems with rank 2.

Table 2 shows the value Score(k) for significant hits at different time intervals for a 50:50 mix of a Hexokinase (KIBYHA) and human α-lactalbumin (LAHU) sample run on an LC-IT-TOF (Liquid Chromatography - Ion Trap - Time of Flight) Mass Spectrometer. The sample is separated into fractions based upon retention time by the liquid chromatography column. The first column is the fraction in retention time minutes. There is a known contaminant fructose-bisphosphate aldolase (yeast) in the sample.

TABLE 2

The method correctly identifies hexokinase A, α-lactalbumin and fructose- bisphosphate aldolase. Although Hexokinase B is incorrectly identified it is a protein that is very similar to hexokinase A.

Table 3 shows the value Score(k) for significant hits at different time intervals for a 50:50 mix of Lysozyme (Chicken) and Glycerokinase (E. CoIi) sample run on an LC- IT-TOF Mass Spectrometer. There is again a known contaminant fructose- bisphosphate aldolase (yeast) in the sample.

TABLE 3

The method correctly identifies Lysozyme and Glycerol kinase.

The results of Table 2 and Table 3 were produced with a search using the correct taxonomies which allows increased search speed as less substances are required to be searched.

Whilst the above has been generally described with reference to processing peptides and polypeptides, it will be appreciated that the systems and methods described are equally applicable to any substance for which a fragmentation can be predicted. In particular, the systems and methods described above can be used in the determination of the presence or absence of carbohydrates, metabolytes, lipids, proteoglycans or any other suitable substance in a sample.

Claims

CLAIMS:

1. A method of determining an indication of the presence or absence of a substance in a sample comprising:

receiving data generated from said sample, the received data comprising a plurality of data points, each data point having an associated value indicative of a mass or mass to charge ratio;

obtaining theoretical data points associated with said substance;

for each of said plurality of data points, determining a probabilistic score, said probabilistic score being determined based upon a range of values, said range being defined relative to said associated value;

processing said theoretical data points and said plurality of data points to determine correspondence between said theoretical data points and said plurality of data points;

determining a strength of said correspondence, said strength of said correspondence being based upon said probabilistic scores; and

determining the presence or absence of said substance based upon said strength of said correspondence.

2. A method according to claim 1 , wherein each data point indicates an abundance of chemical species having the associated mass or mass to charge ratio.

3. A method according to claim 1 or 2, wherein for each of said plurality of data points, said range is a range centred upon said associated value.

4. A method according to any preceding claim, wherein each of said theoretical data points has an associated mass or mass to charge ratio.

5. A method according to any preceding claim, wherein, for each of said data points, said probabilistic score is based upon a number of theoretical data points having an associated value within said range.

6. A method according to claim 5, wherein, for each of said data points, said probabilistic score is based upon the number of theoretical data points associated with said substance.

7. A method according to any preceding claim, wherein said theoretical data points are associated with a class of substances of which said substance is a member.

8. A method according to any preceding claim, wherein each theoretical data point specifies a probabilistic value for the associated value.

9. A method according to any preceding claim wherein processing said theoretical data points and said plurality of data points to determine correspondence between said theoretical data points and said plurality of data points comprises: determining a shared peak count, said shared peak count indicating a number of said plurality of data points having an associated value that corresponds to an associated value associated with one of said theoretical data points.

10. A method according to claim 9, wherein determining a strength of correspondence comprises generating a further probabilistic score based upon said probabilistic scores associated with each of said data points and said shared peak count.

1 1. A method according to claim 10, wherein generating said further probabilistic score comprises multiplying said probabilistic scores associated with each of said data points.

12. A method according to claim 11 , further comprising:

enumerating a plurality of combinations; and

summing or multiplying said probabilistic scores based upon said enumerated combinations.

13. A method according to claim 12, wherein enumerating said plurality of combinations comprises:

storing a first combination; and

performing one or more bit-shift operations to enumerate others of the combinations.

14. A method according to any one of claims 10 to 13, wherein said further probabilistic score indicates a likelihood that a number of theoretical data points indicated by said shared peak count have associated values which correspond to associated values of said received data points.

15. A method according to any one of claims 10 to 13, wherein said further probabilistic score indicates a likelihood that a number of theoretical data points equal to or greater than said shared peak count by chance have associated values which correspond to associated values of said received data points.

16. A method according to any one of claims 10 to 13, wherein said further probabilistic score indicates a likelihood that a first plurality of said data points have an associated value that correspond to an associated value associated with one of said theoretical data points and others of said plurality of data points have associated values that do not correspond to an associated value associated with one of said theoretical data points.

17. A method according to any preceding claim, wherein said received data is spectral data.

18. A method according to claim 17, wherein each data point in said received data is a peak in said spectral data.

19. A method according to any preceding claim, wherein said sample comprises at least one chemical entity selected from the group consisting of proteins, peptides, carbohydrates and metabolites.

20. A method according to any preceding claim, wherein said substance is selected from the group consisting of proteins, peptides, carbohydrates and metabolites.

21. A method according to any preceding claim, wherein said theoretical data associated with said substance comprises spectral data.

22. A method according to claim 21 , wherein each theoretical data point is a peak in said spectral data.

23. A method according to any preceding claim, wherein at least part of the method is carried out using a field-programmable gate array.

24. A method according to claim 23 as dependent upon claim 12 or 13, wherein enumerating said plurality of combinations is carried out using a field-programmable gate array.

25. A computer program comprising computer readable instructions configured to cause a computer to carry out a method according to any one of claims 1 to 24.

26. A computer readable medium carrying a computer program according to claim 25.

27. A computer apparatus for determining the presence or absence of a substance in a sample comprising:

a memory storing processor readable instructions; and

a processor arranged to read and execute instructions stored in said memory; wherein said processor readable instructions comprise instructions arranged to control the computer to carry out a method according to any one of claims 1 to 24.

28. Apparatus for determining the presence or absence of a substance in a sample comprising: means for receiving data generated from said sample, the received data comprising a plurality of data points, each data point having an associated value; means for obtaining theoretical data points associated with said substance; means for, for each of said plurality of data points, determining a probabilistic score, said probabilistic score being determined based upon a range of values, said range being defined relative to said associated value;

means for processing said theoretical data points and said plurality of data points to determine correspondence between said theoretical data points and said plurality of data points;

means for determining a strength of said correspondence, said strength of said correspondence being based upon said probabilistic scores; and

means for determining the presence or absence of said substance based upon said strength of said correspondence.

29. A method for processing spectral data, said processing comprising a plurality of processing phases, the method comprising:

executing processing of a first processing phase on a first processing element; and

executing processing of a second processing phase on a second processing element.

30. A method according to claim 29 wherein said spectral data is processed by said first processing phase at a first time; and

said spectral data is processed by said second processing phase at a second time subsequent to said first time.

31. A method according to claim 29 wherein said first processing phase processes first spectral data and said second processing phase processes second spectral data, said second spectral data being different to said first spectral data.

32. A method according to any one of claims 29 to 31 wherein said processing comprises carrying out a method according to any one of claims 1 to 23.

33. A mass spectrometry data analyser for processing spectral data, said processing comprising a plurality of processing phases, the mass spectrometry data analyser comprising:

a first processor arranged to execute processing of a first processing phase; and

a second processor arranged to execute processing of a second processing phase.

34. A method of processing a sample to determine an indication of the presence or absence of a substance in said sample, said substance having associated theoretical data, comprising:

receiving a first data value indicating a first degree of correspondence between said theoretical data and data generated from said sample at a first time; receiving a second data value indicating a second degree of correspondence between said theoretical data and data generated from said sample at a second time; and

determining an indication of the presence or absence of said substance in said sample based upon said first degree of correspondence and said second degree of correspondence.

35. A method according to claim 34, wherein determining an indication of the presence or absence of said substance in said sample comprises:

determining that said substance is present in said sample if both of said first data value and said second data value exceed a predetermined threshold.

36. A method according to any one of claims 34 to 35, wherein determining an indication of the presence or absence of said substance in said sample is based upon a difference between said first and second data values.

37. A method according to any one of claims 34 to 36, wherein determining an indication of the presence or absence of said substance in said sample is based upon a rate of change indicated by said first and second data values.

38. A method according to any one of claims 34 to 37, further comprising:

processing said first and second data values to determine whether to carry out processing to receive a third data value indicating a third degree of correspondence between at least one of said theoretical data points and data generated from said sample at a third time.

39. A method according to claim 38, wherein determining whether to carry out processing to receive a third data value is based upon rate of change between said first and second data values.

40. A method according to claim 38, wherein determining whether to carry out processing to receive a third data value is based upon a relationship between said first and second data values and a threshold value.

41. A method according to any one of claims 38 to 40, wherein:

if it is determined not to carry out processing to receive a third data value, selecting a further substance and further processing said sample to determine an indication of the presence or absence of said further substance in said sample.

42. A method according to any one of claims 34 to 41, wherein said data generated from said sample is spectral data.

43. A method according to claim 42, wherein said data generated from said sample comprises a plurality of peaks.

44. A method according to any one of claims 34 to 43, wherein said sample comprises at least one chemical entity selected from the group consisting of proteins, glycoproteins, peptides, carbohydrates and metabolites.

45. A method according to any one of claims 34 to 44, wherein said substance is selected from the group consisting of proteins, glycoproteins, peptides, carbohydrates and metabolites.

46. A method according to any one of claims 34 to 45, wherein said theoretical data associated with said substance comprises spectral data.

47. A method according to any one of claims 34 to 46, further comprising graphically displaying said first data value and said second data value to a user.

48. A method according to any one of claims 34 to 47 wherein said first and second data values are determined according to any one of claims 1 to 24.

49. A method according to any one of claims 34 to 48, wherein said first data value is generated by first processing and said second data value is generated by said first processing.

50. A method for analysing the composition of a sample, the method comprising: generating a first spectrum based upon said sample;

selecting a peak in said first spectrum to form the basis for further analysis; generating a second spectrum based upon chemical entities represented by said selected peak;

processing said second spectrum to determine an indication of the presence or absence of a substance in said sample by carrying out a method according to any one of claims 45 to 49.

51. A method according to claim 50 wherein if said indication satisfies a first criterion the method further comprises;

selecting a second peak in said first spectrum;

generating a further second spectrum based upon chemical entities represented by said second peak; and

processing said further second spectrum to determine an indication of the presence or absence of a substance in said sample.

52. A method according to claim 50 wherein if said indication satisfies a second criterion, the method further comprises: generating a further second spectrum based upon chemical entities represented by said selected peak; and

53. A method according to claim 50, wherein if said indication satisfies a third criterion, the method further comprises:

selecting a third peak in said second spectrum to form the basis for further analysis;

generating a third spectrum based upon chemical entities represented by said third peak; and

processing said third spectrum to determine an indication of the presence or absence of a substance in said sample.

54. A method according to any one of claims 50 to 53, further comprising graphically displaying said selected peaks to a user.

55. A mass spectrometry system comprising:

a mass spectrometry apparatus suitable for performing tandem mass spectrometry;

a memory storing processor readable instructions; and

a processor arranged to read and execute instructions stored in said memory; wherein said processor readable instructions comprise instructions arranged to control the computer to carry out a method according to any one of claims 1 to 24 or 29 to 32 or 34 to 54.