Identifying and quantifying list of probable principal IONS in chromatographic dimension present in a material

Abstract

The present subject matter is related to identification / extraction and quantification of mass separated chromatographic peaks obtained from chromatography coupled mass spectrometry instruments This includes both liquid and gas chromatography as well as high resolution and triple quadrupole type mass analyzers. This also includes the case when these are operated in conjunction with ion mobility separation devices. The raw data generated by the instrument is first converted into an open source format such as the mzML format, obvious noise is removed by applying ad hoc filters, mass scans are centroided, regions of interest (ROI) are built in Retention Time dimension, ROIs are binned into 3 or more buckets, longer ROIs are broken into special smaller ROIs using overlapping and sliding window approach, ROIs are classified as peak or noise, peak boundaries identified if classified as peak, longer ROIs formed from special smaller ROIs, and chromatography coupled mass spectrometry data Features are quantified and reported in Retention Time dimension. The ROI classification and peak boundary detection steps make use of specially trained deep learning methods.

Description

IDENTIFYING AND QUANTIFYING LIST OF PROBABLE PRINCIPAL IONS IN CHROMATOGRAPHIC DIMENSION PRESENT IN A MATERIALTECHNICAL FIELD

[0001] The present subject matter, in general, relates to mass spectrometry and more particularly to data in retention time vis-a-vis intensity domain of mass spectrometry instruments.BACKGROUND OF THE INVENTION

[0002] Mass spectrometry (MS) is an analytical technique that is used to measure the mass-to-charge ratio of ions. The results are presented as a mass spectrum, a plot of intensity as a function of the mass-to-charge ratio. Mass spectrometry is used in many different fields and is applied to pure samples as well as complex mixtures. In a typical MS procedure, a sample, which may be solid, liquid, or gaseous, is ionized, for example by bombarding it with a beam of electrons. This may cause some of the sample's molecules to break up into charged fragments or simply become charged without fragmenting. These are commonly referred to as fragment ions and precursor ions, respectively and collectively as the fragmentation pattern of the molecule.

[0003] These ions are then separated according to their mass-to-charge ratio, for example by accelerating them and subjecting them to an electric or magnetic field: ions of the same mass-to-charge ratio will undergo the same amount of deflection. Results are displayed as spectra of the signal intensity of detected ions as a function of the mass-to-charge ratio (denoted as m / z). The molecules in the sample can be identified by matching the m / z values with those of the known molecules or through a characteristic fragmentation pattern.

[0004] Mass spectrometry has become one of the methods of choice in the analytical field due to its high sensitivity and selectivity to retrieve structural information allowing the unequivocal identification of compounds.

[0005] Mass spectrometers are designed to acquire mass scans that can be viewed as two dimensional charts of intensity versus the mass-to-charge ratio (denoted as m / z) of the ionsgenerated by an ion source. When connected to a liquid chromatography (LC), gas chromatography (GC), capillary electrophoresis (CE) or ion mobility separation (IMS) device, the mass scans can be collected continually and as a function of chromatographic retention time or a measure of ion mobility, thereby making this a three-dimensional data. Additionally, attachment of two of these devices to an HR-MS instrument has the potential to generate four-dimensional data. Inspite of various benefits mass spectrometry holds, there are shortcomings in terms of accuracy in certain scenarios.

[0006] Chromatography, including both liquid (LC) and gas chromatography (GC), is an analytical technique utilized for the separation, identification, and quantification of components within a mixture. It operates on the principle of differential partitioning between a mobile phase and a stationary phase, facilitating the resolution of complex compounds based on their varying affinities. LC employs liquid solvents, while GC utilizes inert gasses, each tailored to specific analytical needs. This technique's precision and adaptability make it indispensable in diverse fields such as pharmaceuticals, environmental monitoring, and chemical research.

[0007] Eurther, the coupling of mass spectrometry with liquid (LCMS) or gas (GCMS) represents a significant advancement in analytical technology, addressing the need for enhanced sensitivity and specificity in compound analysis. This integrated approach leverages the separation capabilities of chromatography with the powerful detection and structural elucidation offered by mass spectrometry. The synergy between these techniques facilitates the accurate identification and quantification of complex mixtures, even at trace levels.

[0008] Chromatography coupled mass spectrometer generates complex 3D data where intensity is a function of retention time (RT) and mass-by-charge (m / z). Moreover, it is riddled with noise and redundant signals. Consequently, processing such data is non-trivial. Several open-source software are available for data analysis, such as MS-Dial, XCMS, MZmine, etc. However, they primarily use rudimentary signal processing algorithms and sorely lack precision. Sometimes, the multiple dimension data produced is generally referred to as mass spectrometry data, while it holds chromatography and ion mobility dimension data as well.

[0009] A Canadian patent CA2465297C proposes a method of identifying molecules of biological origin. The molecules are identified on the basis of the accurately determined mass to charge ratio of the molecules and an additional physico-chemical property such as elutiontime or charge state. Further physico-chemical properties may be used. The experimentally determined accurate mass and physico-chemical properties can then be compared with a look-up table of information. The look-up table may be generated or physico-chemical properties of data in a conventional database may be calculated. The ability to recognise and preferably identify the same molecules in two different samples may be used to determine whether a particular biological molecule has been expressed differently in an experimental sample relative to a control sample. Centroiding is used to determine the Intensity Value for all isotopes of all charge-states of any ion exceeding the minimum threshold for ion detection.

[0010] However, there is a lack of accurate extraction and quantification of mass spectrometry data in Retention Time dimension.SUMMARY OF THE INVENTION

[0011] The present invention discloses deep learning-based extraction and quantification of mass-separated chromatographic peaks from the data collected using chromatography coupled mass spectrometry instruments. In one embodiment, this may be done after centroiding the mass scan data.

[0012] For a specific mass by charge value being shared by more than one compound / element, being able to distinguish between one compound / element from the other is an aspect of this invention. The present invention helps identify small and large molecules with better accuracy using chromatography coupled mass spectrometry.

[0013] A compound can be characterized by a combination of the values of retention time and mass to charge ratio. However, as a limitation of the analytical measurement technique, a dispersion is observed around these characteristic values thus resulting in gaussian-like peaks in intensity vs. mass / charge ratio and intensity vs. retention time. As one aspect, the present invention uses intensity weighted averaging of the dispersion in retention time to accurately calculate the associated properties.

[0014] In one embodiment, the method makes use of convolutional neural networks (CNN). CNN may be considered as a type of deep learning method that imitates the way humans gain certain types of knowledge. Deep Learning has proved to be a very powerful tool because of its ability to handle large amounts of data. Deep learning methods are stacked in ahierarchy of increasing complexity and abstraction. One of the most popular deep neural networks is Convolutional Neural Networks.

[0015] In the process, a Convolutional Neural Network (ConvNet / CNN) may take in an input image or tensors, assign importance (learnable weights and biases) to various aspects / objects in the image and be able to differentiate one from the other. Further methods for further noise reduction and elimination have been proposed.

[0016] Accordingly, the present invention provides a method of identifying and quantifying a list of probable principal ions in chromatographic dimension, present in a material, the method comprising the steps of receiving a chromatography coupled mass spectrometry data for multiple retention times . Further, building Regions of interest (ROIs) in retention time vs. intensity dimensions from said chromatography coupled mass spectrometry data may take place. Further, binning the Regions of interest (ROIs), into multiple bins, depending on the size of the ROIs is provided. To identify the ROIs and rightly identify their qualities, deep learning techniques may be used. Accordingly, the further step involves, applying a specially trained database of a first deep learning method to classify ROIs as peak or noise depending on the size of the ROI . Applying a specially trained database of a second deep learning method to mark peak boundaries, depending on the size of the ROI Quantifying the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data.

[0017] Further, another aspect of the present disclosure is providing the method of identifying and quantifying a list of probable principal ions in chromatographic dimension, wherein the ROIs are binned into three bins namely less than a lower threshold called small ROIs, between lower threshold and a higher threshold called medium ROIs, and longer than a higher threshold are considered long ROIs.

[0018] Another aspect of the disclosure herein is the method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 2, further comprising the steps of converting long ROI data into one or more special ROI of length equal to the said higher threshold and binning said ROIs into respective bin, before the step of applying a specially trained database of a first deep learning method to classify ROIs as peak or noise, collating the peak boundary data obtained from above step for special ROIs to mark peak boundaries for the large ROIs formed out of a combination of such special ROIs,before the step of quantifying the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data.

[0019] Another aspect of the disclosure is converting long ROI into special ROI is done using Interpolation, segmenting, sliding or overlapping windows.

[0020] Still another aspect of the disclosure is converting ROI with length lower than lower threshold into ROI with length equal to lower threshold using interpolation or zero padding, and adding each of said ROI to the first subset. This helps in having a standard size of ROIs to deal with, converting ROI with length between lower threshold and the upper threshold into the ROI with length equal to the upper threshold using interpolation or zero padding, and adding each of said ROI to the second subset.

[0021] Still another aspect of the disclosure, is that the mass spectrometry data dealt with is a three dimensional or a four dimensional data comprising at least: m / z, retention time, and intensity values. Further, the disclosure provides for dealing with the input mass spectrometry data that is centroided data. Also, the possibility of dealing with non centroided data is available with reduced accuracy.

[0022] Another aspect of the disclosure a region of interest is a range of retention time, such that the intensity values corresponding to the retention time values are either continuously non-zero with zero values, if any, being isolated by being preceded and succeeded by a non-zero value and the ROI having an ROI boundary of a predefined consecutive zero intensity values. In various embodiments, smoothening of data that are classified as peaks may also take place.

[0023] Another aspect of the disclosure involves steps of creating a data for the specially trained database of first deep learning method which comprises steps of fetching a set of training-chromatography-coupled-mass-spectrometry-data-files as pre-identified datafiles, identifying all the Regions of Interest (ROI) within one selected datafile from the set of training-chromatography-coupled-mass-spectrometry-datafiles choosing a predefined number of small ROIs from above formed ROIs and labeling them as first subset. Further choosing a predefined number of medium ROI, from above formed ROIs and labeling them as second subset, then converting ROI with length lower than lower threshold into ROI with length equal to lower threshold using interpolation or zero padding, and adding each of said ROI to the first subset, then converting ROI with length between lower thresholdand the upper threshold into the ROI with length equal to the upper threshold using interpolation or zero padding, and adding each of said ROI to the second subset. The steps are repeated for each pre- identified datafile from the set of training-chromatography-coupled-mass-spectrometry-datafiles. Lastly, identifying the ROIs as containing at least one true peak, as such, based on identification by a user and classifying them as peak data of first subset and peak data of second subset, otherwise labeling said ROIs as noise data of first subset and noise data of second subset.

[0024] Yet another aspect of the disclosure selection from multiple training regions of interest (training ROIs), is done from a spread out range of retention time values in each of the said pre-identified datafile.

[0025] Another aspect of the disclosure involves creating a data for the specially trained database of second deep learning method which comprises steps of fetching the peak data of first subset and peak data of second subset containing at least one true peak, marking the start and end boundary of all the peaks within the confines of the ROI by a user and marking the output file as processed peak data of first subset and processed peak data of second subset, respectively.

[0026] Still another aspect of the disclosure, the first deep learning method comprises the step of fetching the training mass spectrometry data file from in-house and open sources. The pre-identified set of data files is acquired from instruments of more than one vendor of chromatography and mass spectrometry instruments.

[0027] Another aspect of the disclosure deals with providing a non-transitory computer readable storage medium to tangibly store instructions, which when executed by a computer, cause the computer to receive a chromatography coupled mass spectrometry data for multiple retention times. Further, build Regions of interest (ROIs) in retention time vs. intensity dimensions from said chromatography coupled mass spectrometry data. Further, Binning the Regions of interest (ROIs), into multiple bins, depending on the size of the ROIs takes place. Further, instructions cause the computer to apply a specially trained database of a first deep learning method to classify ROIs as peak or noise depending on the size of the ROI. Further instructions include applying a specially trained database of a second deep learning method to mark peak boundaries, depending on the size of the ROI, and quantify the properties of the peaks representing a probable principal ion in the chromatographycoupled mass spectrometry data.

[0028] Various embodiments herein also provide a system for identifying and quantifying list of probable principal ions in chromatographic dimension, present in a material, comprising a region of interest(ROI) identification module, configured to identify a region of interest(ROI) as long ROI, normal ROI or small ROI, wherein, an ROI of size larger than a higher threshold is a long ROI, an ROI of size smaller than the lower threshold is a small ROI and an ROI of size between lower threshold and higher threshold is medium ROI. Further, an ROI fragmentation module is provided configured to break each of the long ROI into medium ROI. For allowing the intelligence of experts to various users, a first deep learning module, is provided, that is configured to check if said ROI contains at least one true peak, as that caused due to the presence of a substance in the sample and not because of an instrument noise, wherein the checking is based upon the specially trained database of first deep learning method. Similarly, a second deep learning module is also provided, that is configured to mark a peak boundary of the peak based on a second deep learning method, wherein the marking is based upon the specially trained database of the second deep learning method. Further, a peak quantification module is provided that is configured to quantify the properties of the peak identified above.BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS

[0029] A clear understanding of the key features of the invention summarized may be had by reference to the appended drawings and their respective captions, read with the description, which illustrate the deep learning-based extraction and quantification of mass spectrometry data, although it will be understood that such drawings depict preferred embodiments herein and, therefore, are not to be considered as limiting its scope with regard to other embodiments which the invention is capable of contemplating.

[0030] In the following description, for the purpose of explanation, specific details are set forth in order to provide an understanding of the invention. It will be apparent, however, to one skilled in the art that the invention may be practiced without these details. One skilled in the art will recognize that embodiments of the present invention, some of which are described below, may be incorporated into multiple mass spectrometry techniques. Structures and devices as shown in the block diagram are illustrative of exemplary embodiments of theinvention and are meant to avoid obscuring the invention. Furthermore, connections between components within the figures are not intended to be limited to direct connections.

[0031] Reference in the specification to “one embodiment” or “an embodiment” means that a particular feature, structure, characteristic, or function described in connection with the embodiment is included in at least one embodiment of the invention. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment. The headings and subheadings used in the document are not intended to limit the content therein to the respective heading or subheading. Rather, they are used to help readers navigate and prevent obscuring the invention. Accordingly:

[0032] Figure 1 depicts a flow chart of a method for identifying and quantifying peak properties of mass spectrometry data in retention time vs intensity obtained from chromatography coupled mass spectrometry instruments, using deep learning as per an embodiment herein.

[0033] Figure 2a illustrates a mass trace as per an embodiment herein and Figure 2b illustrates the corresponding chromatographic peak as per an embodiment herein.

[0034] Figure 3a illustrates an example of an arbitrarily chosen, real life long Region of Interest (ROI) in the retention time vs intensity plot as per an embodiment herein, and Figure 3b, 3c, 3d and 3e illustrate the formation of special small ROI’s using sliding and overlapping window approach, as per an embodiment herein.

[0035] Figure 4 illustrates graphs assisting in understanding of the method of classifying the ROIs into peak and noise, using deep learning methods as per an embodiment herein.

[0036] Figure 5a and 5b illustrate the smoothening of regions of interest (retention time vs intensity plot) that have been classified as peaks in the previous step as per embodiment herein.

[0037] Figure 6 (collectively 6a, 6b, 6c and 6d) illustrates graphs assisting in understanding of the method of identification of peak boundaries in ROIs by using a deep learning-based method, as per an embodiment herein.

[0038] Figure 7 illustrates a representative block diagram of a system for identifying and quantifying a list of probable principal ions in chromatographic dimension, present in a material, as per an embodiment herein.DETAILED DESCRIPTION OF THE INVENTION

[0039] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well as the singular forms, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0040] Unless, otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one having ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein. Further, references to deep-learning may be understood to refer to machine learning as well.

[0041] In describing the invention, it will be understood that a number of techniques and steps are disclosed. Each of these has individual benefits and each can also be used in conjunction with one or more, or in some cases all, of the other disclosed techniques. Accordingly, for the sake of clarity, this description will refrain from repeating every possible combination of the individual steps in an unnecessary fashion.

[0042] Various embodiments herein disclose machine learning-based extraction and quantification of mass spectrometry data. This data may be collected using chromatography coupled mass spectrometry instruments. Also, reference to the deep learning-based identifying and quantifying peak properties of mass spectrometry data in retention time dimension, as may be understood by a person skilled in the art or otherwise as described herein.

[0043] Quantifying the properties of the peaks in the mass spectrometry data in Retention time dimension is provided. The properties of the peak may comprise the m / z value associated with the peak, the retention time value associated with the peak, the area under the peak, and the Signal to Noise Ratio (SNR) for the peak.

[0044] The embodiments herein also explain the use of certain deep learning techniques for peakidentification in the retention time vs. intensity plots obtained from mass separated chromatogram data and specially designed feature vectors as input to the method to classify peak and noise in chromatograms. The deep learning technique uses a feature vector (X) as an input and predicts the class (Y). The feature vector of an ROI can be the intensity vector, normalized intensity vector (values normalized between 0 and 1), a suitably smoothened, normalized intensity vector or a derivative vector of one of the above vectors, typically obtained by a finite difference method. A person skilled in the art may realize the use of alternative feature vectors with corresponding changes to suit the working of the embodiments herein. The input X can also be a tuple based on a suitable combination of the feature vectors described above. X can have the size of n x 1, n x 2 or n x 3 depending on whether the input comprises one, two or three feature vectors, respectively, from the ones described above. The output Y is a class label such as peak or noise.

[0045] The present invention is also related to the detection of peaks and their boundaries in the retention time vs. intensity plots obtained from mass separated chromatograms and removal of noise by using deep learning techniques. The deep learning method applies statistically driven threshold to set values below the threshold to be zero, breaks the long vectors of intensity into smaller vectors or regions of interest (ROI), extracts features of the ROIs, submits them to a specially trained deep learning method, predicts peaks and their boundaries and calculates the median, mean or weighted average m / z, median, mean or weighted average retention time area under the curve for each peak and, signal to noise ratio. The deep learning tool is trained to perform the task of detecting peaks and their boundaries. The peaks are of variable heights and widths. The present invention uses the retention time separated mass scans to form regions of interest in the retention time vs. intensity plot. Further, the intensity is normalized in the range of 0 and 1, thereby zooming along the y-axis. A deep learning method classifies the Region of interest (ROIs) into peak and noise. Another deep learning method then marks the peak boundaries and allowing the calculation of (i) the median, mean or weighted average m / z of the peak, (ii) the median, mean, or weighted average retention time of the peak, (iii) the peak area as area under the curve between the peak boundaries and in terms of intensity*minutes, and (iv) Signal to Noise Ratio (SNR).

[0046] The deep learning methods use the normalized intensity vector and a vector of firstderivatives of the intensity vector of an ROI as inputs and provide classification of the ROI as peak or noise and further provide peak boundaries if classified as a peak. In one embodiment, the method may be primarily based on convolutional neural networks (CNN). A Convolutional Neural Network (ConvNet / CNN) is a Deep Learning method which can take in an input image, assign importance (learnable weights and biases) to various aspects / objects in the image and be able to differentiate one from the other.

[0047] The deep learning-based method is trained by using real examples of ROIs that are labeled by human experts. Labeled data is not available in the public domain and needs to be generated specially by using real life chromatography coupled mass spectrometry data. The ROIs are generated from the real life, chromatography coupled mass spectrometry data available in the public domain that has been generated by instruments from various vendors.

[0048] The list of peaks that represent probable elements / compounds present in the sample and the properties are calculated that will be utilized by analysts for detailed analysis of the study. The retention time values and peak areas are then processed further to build peaks in the drift time (DT) dimension if the data was acquired in an ion-mobility high-resolution mass spectrometry (HR-MS) instrument. The presence of noisy peaks data results in a noisy inaccurate analysis, hence, affecting the purpose of study. Thus, the present invention significantly reduces the noisy peaks in the data thereby significantly improving the quality of the analysis of biological molecules. This in turn improves the outcome of the results from the study.

[0049] Figure 1 of the present embodiment explains the flow chart of deep learning-based identification and quantification of mass spectrometry data, as per an embodiment herein. In an embodiment, data needs to be centroided if data is acquired in Profile mode. Data can be acquired in centroided mode in the HR-MS instrument and if the data is acquired in profile mode, it can be centroided by various softwares and algorithms present.

[0050] In one embodiment, the steps may include receiving a chromatography coupled mass spectrometry data for multiple retention times, building Regions of interest (ROIs) in retention time vs. intensity dimensions from said chromatography coupled mass spectrometry data, binning the Regions of interest (ROIs), into multiple bins, depending on the size of the ROIs, applying a specially trained database of a first deep learning method to classify ROIs as peak or noise depending on the size of the ROI, applying a speciallytrained database of a second deep learning method to mark peak boundaries, depending on the size of the ROI further quantifying the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data.

[0051] In an embodiment, the raw data or analytical data that is received from the high-resolution mass spectrometer is first converted into an open format such as the mzML format. Centroiding of data is done if the data is acquired in profile mode, else centroid data as obtained may be used.

[0052] In one embodiment, building of regions of interest (ROIs) from retention time separated mass scans from dynamically chosen m / z values takes place and the retention time is plotted against the intensity. Building Region of Interest (ROIs) in retention time vs intensity dimensions from centroided data may take place

[0102] , This may be further understood with reference to Figure 2a, plotting Retention time for a small range of m / z value, while the drift time (if present) is constant. The various dots may be color coded to indicate the intensity. Further, Figure 2b shows the plot of intensity with retention time. The scans collected at each drift time and various Retention time (RT) points are grouped for this process. If the data is not collected with an ion-mobility mass spectrometer, then all the centroided mass scans form a single group. Each group of scans is processed using the CentWave algorithm. In one embodiment, a CentWave algorithm [https : / / bmcbioinformatics. biomedcentral. com / articles / 10,1186 / 1471-2105 -9-5041 is used to return a list of ROIs. Each ROI formed is 2 dimensional data with RT and intensity for a slice of m / z.

[0053] In one embodiment, the binning

[0103] of the ROIs as per their length into multiple bins may take place. In one exemplary embodiment, there may be three bins, namely: for length < =128 points, 128>length<=256, length > 256. The bin having ROIs with length longer than 256 points may be referred to as “long ROIs” and may need to be further processed. Hence, the higher threshold for considering an ROI a long ROI in this exemplary embodiment would be 256. In another embodiment, different classification of the bins may be undertaken, with a different threshold. Further, the use of additional bins may take place.

[0054] In one embodiment, the ROIs are binned into three bins namely less than a lower threshold called small ROIs, between lower threshold and a higher threshold called medium ROIs, and longer than a higher threshold are considered long ROIs.

[0055] In one embodiment, the steps involved would include converting long ROI data into one or more special ROI of length equal to the said higher threshold and binning said ROIs into respective bin, before the step of applying a specially trained database of a first deep learning method to classify ROIs as peak or noise, collating the peak boundary data obtained from above step for special ROIs to mark peak boundaries for the large ROIs formed out of a combination of such special ROIs, before the step of quantifying the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data.

[0056] In one embodiment, the ROIs that are of size k (k being less than n) can be forced to be of the target size either by zero padding or padding with very low intensity of noise (upper limit of noise padded can be taken as 1% of maximum intensity of ROI) or by interpolating and resampling. Where the target size may be the size of the window that the deep learning model is able to process.

[0057] In one embodiment, the method of identifying and quantifying a list of probable principal ions in chromatographic dimension, wherein converting long ROI into special ROI may be done using Interpolation, segmenting, sliding or overlapping windows.

[0058] In one embodiment, the further steps may include converting ROI with length lower than lower threshold into ROI with length equal to lower threshold using interpolation or zero padding, and adding each of said ROI to the first subset, converting ROI with length between lower threshold and the upper threshold into the ROI with length equal to the upper threshold using interpolation or zero padding, and adding each of said ROI to the second subset.

[0059] In one embodiment, the mass spectrometry data is a three dimensional or a four dimensional data comprising at least: m / z, retention time, and intensity values. In one embodiment, the mass spectrometry data that is used may be centroided data. In one embodiment, the multiple bins are of the sizes: less than 128, 128 to 256, and greater than 256 data points.

[0060] On breaking the longer ROIs, smaller ROIs of length <= 256 by using a sliding and overlapping window approach may be obtained. The ROIs thus obtained may be referred to as “special small ROIs”. Thus, ROIs that are smaller than the target size can be forced to be of the target size either by zero padding or padding with very low intensity of noise (upperlimit of noise padded can be taken as 1% of maximum intensity of ROI) or by interpolating and resampling. As may be seen in reference to the exemplary scenario in Figure 3a the intensity vs retention time plot of a range of retention time 0 to 650 is split into 4 windows in Figure 3b having a range of 0 to 256, 128 to 384, 256 to 512, and 384 to 650. In another embodiment, the size of the windows may vary and may depend upon the deep learning model used and the target size that the deep learning model is able to process.

[0061] In one embodiment, the further steps may involve applying a specially trained first deep learning model to classify ROIs as peak and noise. In another embodiment, additional class labels may be used. For example, ternary, quaternary etc. or in general, multivalent labels may be used. For example, we can choose the output variable to be ternary where the 3 options are 1) Noise, 2) Peak and 3) Borderline cases (ROIs which have some gaussian characteristics but have, say low SNR).

[0062] The normalized intensity vector along with certain derived values are then submitted as features to apply

[0104] specially trained deep learning model to classify the ROI as peak or noise and to correctly identify the peak boundaries. The specifically used deep learning model Convolutional Neural Network (CNN) is used for classification of ROIs and for detection of peak boundaries. Region of Interests (ROIs) formed in the previous steps may be given as an input to this step of ETL (Extract Transform Load) data pipeline. The input to deep learning model is intensity vector, its first derivative and other vectors calculated from intensity vector. Convolutional Neural Network may be used to classify peaks (good quality signal) and noise (artifact). The noise is discarded at this step and peaks are fed to the next convolutional neural network (segmentation) for detection of peak boundaries (start and end of peak) after smoothening the intensity vector as depicted.

[0063] In one embodiment, region of interest may be a range of retention time, such that the intensity values corresponding to the retention time values are either continuously non-zero with zero values, if any, being isolated by being preceded and succeeded by a non-zero value and the ROI having an ROI boundary of a predefined consecutive zero intensity values.

[0064] In one embodiment, building Regions of interest accounts for mass tolerance using full width at half maximum (FWHM) of constituent mass centroids.

[0065] Smoothening of data that are classified as peaks (example: Moving Average or anyAlgorithmic smoothening). This step may particularly help remove high frequency noise in the retention time vs intensity plot. This can be understood in detail from Figure 5a that contains retention time vs intensity plot without smoothening and Figure 5b containing the same plot but after removing very high frequency noise. Hence, preparing the data for accurate interpretation helps the second deep learning model to make informed decisions. This may be better understood with reference to Figure 6a depicting an input plot between intensity and retention time and Figure 6d showing corresponding smoothed graph / plot.

[0066] In one embodiment, the further step may include applying

[0105] a specially trained deep learning model to mark peak boundaries as may be seen in Figure 6a is the smoothened ROI, Figure 6b contains the intensity vector, Figure 6c derivative vector and Figure 6d describes the peak boundary detection from second deep learning model. The specific deep learning model used here is Convolutional Neural Network (CNN) that takes the intensity and derivative of intensity as input and output bit mask (a vector that contains 0 or 1) of the same length as intensity vector, which is further processed to conclude the borders of each peak(s) present in a Region of Interest (ROI). These specially trained deep learning models were trained on the real life data annotated by human experts. It may be pertinent to note here that existing solutions in this domain use statistical and mathematical algorithms to detect between noise and peaks. Further, more than 50% of the classified signal as peaks by other software are not good quality signals and thus require human expert intervention to go through thousands of signals. However, the disclosure presented herein for our deep learning models requires little to no human intervention that makes the process of classification of good and bad quality signals. Lack of this functionality of detecting peak boundaries automatically requires a huge amount of human expert time to go through all peaks and mark the peak boundaries.

[0067] In one embodiment, the further step may include using the peak boundaries of special small ROIs (which may be also referred to as special ROIs) formed to mark the peak boundaries of long ROIs from which special small ROIs were formed. This step helps us to remove the limitations of the deep learning models to process the fixed length of input. Further, median, mean or weighted average of Retention time, mean, median and weighted average of m / z value, signal to noise ratio and area under the curve are recorded along with other statistics for each peak. The process is repeated for all the Regions of Interest formed. The stepsshown in 101-106 may be automated on a computing device by writing software programs in languages such as, for example Python or C++. Additionally, public domain libraries such as TensorFlow, Numpy, Scipy, and Keras may be used to deploy the steps of 101-106 on a computer. A person skilled in the art may use other public domain libraries with obvious modifications for suitability with the teachings of the embodiments herein.

[0068] In one embodiment, the first deep learning method comprises the step of fetching the training mass spectrometry data file from in-house and open sources. Further the first pre-identified set of data files is acquired from instruments of more than one vendor of chromatography and mass spectrometry instruments.

[0069] For better understanding an exemplary flow is provided below for creating a data for the specially trained database of first deep learning method:A. Selection of a set of chromatography coupled mass spectrometry datafilesB. Selecting one datafile from the above raw dataset. (Raw dataset = a set of datafiles)C. Identifying all the Regions of Interest (ROI) within the selected datafileD. Choosing a predefined number of ROI, with length < 128, from above formed ROIs.E. Choosing a predefined number of ROI, with 128 < length < 256, from above formed ROIs.F. Ensuring the ROIs are chosen from a spread-out range of retention time values in the selected datafileG. Repeating steps 2 to 6 for all the datafiles in the dataset, to form two training datasets of ROI, one set only contains ROI with length < 128 and another set only contains ROI with 128 < length < 256.H. ROI with length < 128 are forced to be of length 128 via interpolation or zero padding.I. ROI with 128 < length < 256 are forced to be of length 256 via interpolation or zero padding.J. For both datasets, label the ROIs as either GOOD or BAD.Ik. (GOOD = contains at least one true peak. BAD = no true peak = Noise)L. For both datasets, fetching all the ROIs that are good.M. Marking the start and end boundary of all peaks within the confines of the good ROIs by an expert.

[0070] Further, in one exemplary embodiment, the deep learning methods or models for classification (labeling ROI as Good or Bad) and Segmentation (Identifying boundary of all peaks) of ROI, may have following features. Deep learning models may accept an input vector of fixed predefined size. In one scenario these may be two sizes viz. 2 x 128 and 2 x 256. In such scenarios, one may have two classification models (namely Cl 28 and C256) and two Segmentation models (S128 and S256), one for each of the input vector shapes defined above. The two models are individually or independently trained on the two datasets that are prepared as explained in the above section.

[0071] Further identifying the presence of at least one true peak in said ROI, involves steps of normalizing the intensity values in the ROI, by dividing all the intensity values in the ROI by the maximum intensity value in that ROI, ensuring all intensity values in the ROI he between a value zero and a value one and smoothening using a moving average and producing a smoothened intensity vector; taking derivative of the smoothened intensity vector; producing a derivative vector; and concatenating said smoothened intensity vector and said derivative vector to together form an input feature vector for the deep learning method.

[0072] In one embodiment, a non-transitory computer readable storage medium may also be provided, such that it tangibly stores instructions, which when executed by a computer, cause the computer to receive a chromatography coupled mass spectrometry data for multiple retention times that further includes building Regions of interest (ROIs) in retention time vs. intensity dimensions from said chromatography coupled mass spectrometry data, binning the Regions of interest (ROIs), into multiple bins, depending on the size of the ROIs, further the step involves applying a specially trained database of a first deep learning method to classify ROIs as peak or noise depending on the size of the ROI, applying a specially trained database of a second deep learning method to mark peak boundaries, depending on the size of the ROI and then quantifying the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data.

[0073] For better understanding an exemplary flow is provided for identifying and quantifying the list of probable principal ions in chromatographic dimension, present in a material. In this exemplary scenario, the higher threshold may be 256 while the lower threshold is 128. Theflow may be as follows:A. A chromatography coupled mass spectrometry datafile is receivedB. All the ROIs within the said datafile are built / formed.C. The ROIs thus formed are divided into separate bins based on their length into short ROI (length < 128), medium ROI (128 < length < 256), and long ROI (length > 256)D. ROIs are first classified into GOOD or BAD based on classification modelE. The good ROIs are then further passed onto the segmentation model for identification of boundaries of all peaks within the said good ROI.F. All Short ROIs are forced to be of length 128.G. Short ROIs are then processed using Cl 28 and SI 28.H. All Medium ROIs are forced to be of length 256.I. Medium ROIs are processed using C256 and S256.J. Long ROIs are broken into multiple special ROI of at most length 256. All of them are then forced to be of length 256.K. All special ROI are then processed using C256 and S256.L. The output of processing of all special ROI is then collated to obtain the output of the original long ROI.M. Peak properties (area under the curve, median retention time, median mass-by-charge ratio, signal to noise ratio, etc.) of the peaks identified within the ROI are quantified.

[0074] Figure 2 provides examples for formation of regions of interest (ROIs) from retention time separated centroided mass scans. Chromatography coupled mass spectrometry data is sparse and thus ROIs are essentially regions of non-zero intensity values. Certain tunable parameters may be used to build ROIs, which may then be subjected to ad hoc filters to eliminate instances of noise that can be identified readily.

[0075] In one embodiment, centroids from consecutive scans need not have the exact same mass-by-charge value. Therefore, to stitch / group centroids across scans, one must allow for some leeway / tolerance in variation of the mass-by-charge value of centroids across scans. This could be, for example, a fixed value -0.01 Dalton, that is, if two centroids from consecutive scans have a difference in m / z of less than 0.01 Dalton then they can be stitched together to be part of the same ROI. However, if the difference between their m / z is more than 0.01, then they cannot be part of the same ROI. This is exemplified by Figure 2a wherethe vertical width of the rectangle, that spans from approximately 184.96 to 185.00, represents this tolerance. This acceptable value of tolerance is a parameter that is usually decided by the user. This specific parameter is very critical to the quality of the results and a user needs some expertise of the instrument and data to choose a reasonable parameter. Otherwise, the user may end up just trying multiple values and see which gives the best results.

[0076] Further, in an embodiment, a method is provided wherein this value of tolerance can be obtained automatically from the data. Full Width at Half Maximum (FWHM) is a representative of the dispersion of data in spectral dimension and consequently reflects the resolution of the instrument. This resolution is what dictates the tolerance parameter, since it is a tolerance of deviation in the m / z of the centroids. Therefore, we propose using FWHM itself as the tolerance parameter for ROI formation as explained in the previous paragraph.

[0077] Figure 3 of an embodiment provides an illustrative example of a long ROI from an arbitrarily chosen, real life chromatography coupled mass spectrometry data. The example shows a region with several true peaks that are separated by long regions of low but non-zero intensity values. The ROIs that are longer than a certain predetermined target length are then broken using a sliding and overlapping window approach. The special small ROIs formed [Fig. 3b, 3c, 3d, 3e] will be of the length less than or equal to the predetermined length.

[0078] Figure 4 of this embodiment illustrates example ROIs that are classified into peak or noise by using a deep learning-based method. The classification is solely based on the shape of the ROI and not based on the value of intensity. The figure shows a total of 8 ROIs, 4 containing at least 1 species peak and the other 4 comprising pure noise. The method takes the sequences of smoothened, normalized intensity values (normalized between 0 and 1) and its first derivative, stacked into a 2 x k matrix, with k being a predetermined value. ROIs that are shorter than k are padded with zeros or noise (noise will be maximum of 2% of the maximum intensity) on either side. We apply a convolutional neural network with several layers but other architectures of neural networks may be applicable. In an embodiment, the method may be first trained with real life data comprising of ROIs that are labeled by humans as peak or noise. Alternatively, a special purpose application or amethod using computer programs may be used for the said purpose.

[0079] Figure 5 explains the effect of smoothening either by using moving average or any dedicated algorithm to remove the high frequency noise from the peak. Smoothening the peak before giving it as an input to a specially trained second deep learning model helps the model to detect peak boundaries more accurately.

[0080] Figure 6 explains the identification of peak boundaries in ROIs by using a deep learning-based method. Once an ROI is classified as a peak, it may contain one or more peaks and the peak boundaries need to be identified correctly. An ROI [Fig. 6a] obtained from an arbitrarily chosen real life mass separated chromatogram has been used as an illustrative example. The method takes the sequences of smoothened, normalized intensity values (normalized between 0 and 1) [Fig. 6b] and its first derivative [Fig. 6c], stacked into a 2 x k matrix, with k being a predetermined value. ROIs that are shorter than k are padded with zeros on either side while ROIs that are longer than k are interpolated and resampled or forcefully broken using a sliding window approach.

[0081] In one embodiment, converting long ROI into special ROI is done using Interpolation, segmenting, sliding or overlapping windows. We apply a convolutional neural network with several layers but other architectures of neural networks may be applicable. The method is first trained with real life data comprising of ROIs that are labeled by humans as peak or noise along with accurately marked peak boundaries. In the present example, the method detects two peaks [shown in two different shades in Figure 6d] within the ROI and marks the peak boundaries.

[0082] Figure 7 is a block diagram of a system for identifying and quantifying a list of probable principal ions in chromatographic dimension, present in a material as per an embodiment herein.

[0083] In an embodiment, a system identifying and quantifying list of probable principal ions in chromatographic dimension, present in a material, comprising a region of interest (ROI) identification module, configured to identify a region of interest (ROI) as long ROI, normal ROI or small ROI, wherein, an ROI of size larger than a higher threshold is a long ROI, an ROI of size smaller than the lower threshold is a small ROI and an ROI of size between lower threshold and higher threshold is medium ROI, an ROI fragmentation module configured to break each of the long ROI into medium ROI, a first deep learning module,configured to check if said ROI contains at least one true peak, as that caused due to the presence of a substance in the sample and not because of an instrument noise, wherein the checking may be based upon the specially trained database of first deep learning method, a second deep learning module, configured to mark a peak boundary of the peak based on a second deep learning method, wherein the marking may be based upon the specially trained database of second deep learning method and peak quantification module, configured to quantify the properties of the peak identified above.

[0084] Important aspects of this invention are (i) the method to select true peaks in region of interests formed by retention time separated mass scans, (ii) the use of normalized intensity and first derivative of intensities as input vectors for the deep learning methods, (iii) the ability to classify peaks and noise and accurately mark peak boundaries, and (iv) the ability to remove noise with minimal loss of data and without human intervention. The invention also saves time and effort. The disclosure provides method / s, system / s and products that can be used to remove chromatographic noise and automate peak detection in data generated from any combination of chromatography (liquid or gas) and mass spectrometry (high resolution {TOF, Orbitrap} or high sensitivity {Triple Quadrupole}).

[0085] The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and / or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope of the embodiments as described herein and the claims provided.

Claims

I / We Claim:

1. A method of identifying and quantifying list of probable principal ions in chromatographic dimension, present in a material, comprising the steps of: a. Receiving a chromatography coupled mass spectrometry data for multiple retention times (101); b. Building Regions of interest (ROIs) in retention time vs. intensity dimensions from said chromatography coupled mass spectrometry data (102); c. Binning the Regions of interest (ROIs), into multiple bins, depending on the size of the ROIs (103); d. Applying a specially trained database of a first deep learning method to classify ROIs as peak or noise depending on the size of the ROI (104); e. Applying a specially trained database of a second deep learning method to mark peak boundaries, depending on the size of the ROI (105); f. Quantifying the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data (106).

2. The method of identifying and quantifying a list of probable principal ions in chromatographic dimension, wherein the ROIs are binned into three bins namely less than a lower threshold called small ROIs, between lower threshold and a higher threshold called medium ROIs, and longer than a higher threshold are considered long ROIs.

3. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 2, further comprising the steps of: a. converting long ROI data into one or more special ROI of length equal to the said higher threshold and binning said ROIs into respective bin, before the step of applying a specially trained database of a first deep learning method to classify ROIs as peak or noise; and b. collating the peak boundary data obtained from above step for special ROIs to mark peak boundaries for the large ROIs formed out of a combination of such special ROIs, before the step of quantifying the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data.

4. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, wherein: converting long ROI into special ROI is done using Interpolation, segmenting, sliding or overlapping windows.

5. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, further comprising the steps of: a. converting ROI with length lower than lower threshold into ROI with length equal to lower threshold using interpolation or zero padding, and adding each of said ROI to the first subset; b. converting ROI with length between lower threshold and the upper threshold into the ROI with length equal to the upper threshold using interpolation or zero padding, and adding each of said ROI to the second subset;6. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, wherein the mass spectrometry data is a three dimensional or a four dimensional data comprising at least: m / z, retention time, and intensity, values.

7. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, wherein the mass spectrometry data is centroided data.

8. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, wherein the multiple bins are of the sizes: less than 128, 128 to 256, and greater than 256 data points.

9. The method as in claim 3, wherein a region of interest is a range of retention time, such that the intensity values corresponding to the retention time values are either continuously non-zero with zero values, if any, being isolated by being preceded and succeeded by anon-zero value and the ROI having an ROI boundary of a predefined consecutive zero intensity values.

10. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, wherein the Building Regions of interest accounts for mass tolerance using full width at half maximum (FWHM) of constituent mass centroids.

11. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, wherein creating a data for the specially trained database of first deep learning method comprises the steps of: a. fetching a set of training-chromatography-coupled-mass-spectrometry-data-files as a pre-identified datafiles; b. identifying all the Regions of Interest (ROI) within one selected datafile from the set of training-chromatography-coupled-mass-spectrometry-datafiles; c. choosing a predefined number of small ROIs from above formed ROIs and labeling them as first subset; d. choosing a predefined number of medium ROI, from above formed ROIs and labeling them as second subset; e. converting ROI with length lower than lower threshold into ROI with length equal to lower threshold using interpolation or zero padding, and adding each of said ROI to the first subset; f. converting ROI with length between lower threshold and the upper threshold into the ROI with length equal to the upper threshold using interpolation or zero padding, and adding each of said ROI to the second subset; g. repeating the above steps for each pre-identified datafile from the set of training-chromatography-coupled-mass-spectrometry-datafiles; h. identifying the ROIs as containing at least one true peak, as such, based on identification by a user and classifying them as peak data of first subset and peakdata of second subset, otherwise labeling said ROIs as noise data of first subset and noise data of second subset.

12. The method as in claim 11 , wherein the selecting from multiple training regions of interest (training ROIs), is done from a spread out range of retention time values in each of the said pre-identified datafile.

13. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 11, wherein creating a data for the specially trained database of second deep learning method comprising the step of: a. fetching the peak data of first subset and peak data of second subset containing at least one true peak, b. marking the start and end boundary of all the peaks within the confines of the ROI by a user. c. marking the output file as processed peak data of first subset and processed peak data of second subset, respectively.

14. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, wherein the first deep learning method comprises the step of fetching the training mass spectrometry data file from in-house and open sources.

15. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, wherein the first pre-identified set of data files is acquired from instruments of more than one vendor of chromatography and mass spectrometry instruments.

16. The method of identifying and quantifying list of probable principal ions in chromatographic dimension as per claim 3, further comprising the step of, repeating steps 11. a to ll.h till nearly an equal number of examples for both peak and noise labeled ROI are identified.

17. The method as in claim 1, wherein, prior to Identifying the presence of at least one true peak in said ROI, a. normalizing the intensity values in the ROI, by dividing all the intensity values in the ROI by the maximum intensity value in that ROI, ensuring all intensity values in the ROI he between a value zero and a value one; and b. smoothening using a moving average and producing a smoothened intensity vector; taking derivative of the smoothened intensity vector; producing a derivative vector; and concatenating said smoothened intensity vector and said derivative vector to together form an input feature vector for the deep learning method.

18. A non-transitory computer readable storage medium to tangibly store instructions, which when executed by a computer, cause the computer to: a. receiving a chromatography coupled mass spectrometry data for multiple retention times; b. building Regions of interest (ROIs) in retention time vs. intensity dimensions from said chromatography coupled mass spectrometry data; c. binning the Regions of interest (ROIs), into multiple bins, depending on the size of the ROIs; d. applying a specially trained database of a first deep learning method to classify ROIs as peak or noise depending on the size of the ROI; e. applying a specially trained database of a second deep learning method to mark peak boundaries, depending on the size of the ROI; f. quantifying the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data.

19. A computer implemented method for identifying and quantifying list of probable principal ions in chromatographic dimension, present in a material, the method comprising instructions which when executed by the computer cause the computer to: a. receive a chromatography coupled mass spectrometry data for multiple retention times;b. build Regions of interest (ROIs) in retention time vs. intensity dimensions from said chromatography coupled mass spectrometry data; c. Bin the Regions of interest (ROIs), into multiple bins, depending on the size of the ROIs; d. apply a specially trained database of a first deep learning method to classify ROIs as peak or noise depending on the size of the ROI; e. apply a specially trained database of a second deep learning method to mark peak boundaries, depending on the size of the ROI; and f. quantify the properties of the peaks representing a probable principal ion in the chromatography coupled mass spectrometry data.

20. A system identifying and quantifying list of probable principal ions in chromatographic dimension, present in a material, comprising: a. a region of interest (ROI) identification module (701), configured to identify a region of interest (ROI) as long ROI, normal ROI or small ROI, wherein, an ROI of size larger than a higher threshold is a long ROI, an ROI of size smaller than the lower threshold is a small ROI and an ROI of size between lower threshold and higher threshold is medium ROI; b. an ROI fragmentation module (702) configured to break each of the long ROI into medium ROI; c. a first deep learning module (703), configured to check if said ROI contains at least one true peak, as that caused due to the presence of a substance in the sample and not because of an instrument noise, wherein the checking is based upon the specially trained database of first deep learning method; d. a second deep learning module (704), configured to mark a peak boundary of the peak based on a second deep learning method, wherein the marking is based upon the specially trained database of second deep learning method; and e. peak quantification module (705), configured to quantify the properties of the peak identified above.

Images