System and methods for polymer analysis
The method uses machine learning models to analyze nanopore measurement signals, optimizing sequencing efficiency by identifying relevant states for further analysis and reducing non-productive nanopore occupancy, thereby improving throughput in nanopore sequencing.
Patent Information
- Application Number
- GB2025005196
- Authority / Receiving Office
- GB · GB
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-04-07
- Publication Date
- 2026-11-11
AI Technical Summary
Nanopore sequencing technologies face challenges in improving overall throughput and efficiency due to the need to analyze measurement signals that may include irrelevant or non-sequenceable states, leading to inefficient use of computational resources and prolonged nanopore occupancy in non-productive states.
A computer-implemented method using machine learning models or metric-based techniques to analyze measurement signals from nanopores, determining relevant states for further analysis and optimizing computational resources by skipping or unblocking non-relevant states, thereby enhancing sequencing efficiency.
The method improves the sequencing throughput by dedicating computational resources to relevant signal portions, reducing time spent in non-sequenceable states, and optimizing nanopore operation, thus enhancing overall sequencing efficiency.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
Technical Field The present invention relates to techniques for analysing polymers. The disclosure has particular relevance to operating a polymer analysis system. Background Polymers, whether they are synthetic plastics or biopolymers, perform a variety of useful roles in our daily lives. The study and analysis of polymers leads to a better understanding of their properties and to an improved ability to manipulate polymers to our benefit. Many properties of a polymer result from its constitutive sequence of polymer units; varying the sequence may cause the polymer to exhibit a modified property. The sequence of polymer units may regulate the properties of biopolymers such as DNA polynucleotides. For example, DNA encodes the genetic information of an organism in its sequence of four basic nucleotide units. Nucleotide sequences in DNA are transcribed to produce functional RNA, which may facilitate the production of proteins. Sequencing the DNA of individuals is central to our understanding of biological function and disease control. Nanopore sequencing is a widely used technique for sequencing biopolymers such as polynucleotides and peptides. In nanopore sequencing, a nanoscale pore embedded within a membrane serves as a sensor for detecting changes in a property (such as ionic current) resulting from the passage of nucleotide units. The measured changes in the property may be analysed to determine the sequence of nucleotide units in a polynucleotide, for example. Nanopore sequencing enables long sequences (reads) of polynucleotides to be obtained in a single pass, precluding the need to assemble shorter reads, as necessitated by earlier techniques. Moreover, variations in the measured property of the nanopore can be analysed during translocation of the polymer, allowing near-real-time sequencing. However, there remain challenges to improving the overall throughput of nanopore sequencing. The present application addresses some of these and other related challenges. Summary According to an aspect of the present disclosure, there is provided a computer-implemented method of processing a measurement signal comprising measurements of a sample by a sensor unit comprising a nanopore. According to further aspects, there is provided apparatus for carrying out the method, and a computer program product (such as a non-transitory storage medium) comprising machine readable instructions which, when executed by a computer, cause the computer to carry out the method. The method includes analysing the measurement signal using a machine learning model to determine that a portion of the measurement signal is associated with a first state of the nanopore, the first state being one of a plurality of states of the nanopore, and determining whether to further analyse the portion of the measurement signal in dependence on the determination that the portion of the measurement signal is associated with the first state of the nanopore. By determining whether or not to further analyse the portion of the measurement signal based on determining the state of the nanopore associated with the portion, the method allows the polymer analysis system to be operated more efficiently. In particular, the method enables further analysis to be conducted on portions associated with nanopore states that may be most relevant to the analysis objectives. For example, in some cases, the first state may be representative of a polymer being translocated through the nanopore, and the method may include determining to further analyse the portion of the measurement signal. In other cases, the first state may be representative of the nanopore not being occupied by a polymer, and the method may include determining not to further analyse the portion of the measurement signal. In this manner, the method may dedicate computational resources for further analysis of portions associated with a polymer being translocated. Furthermore, the method may also indirectly increase the throughput of the analysis system by reducing the time spent by the nanopore in unwanted states. For example, in some cases, the first state may be representative of a polymer being blocked in the nanopore, and the method may include instructing the sensor unit to unblock the polymer in the nanopore. In this manner, the nanopore may spend less time in a state that does not provide measurements capable of being sequenced. Thus, the overall sequencing efficiency of the polymer analysis system may be enhanced. The polymer to be translocated may be produced by template-directed synthesis of a target nucleic acid. The polymer may comprise a plurality of subunits coupled in a sequence corresponding to a contiguous nucleotide sequence of all or a portion of the target nucleic acid. Each subunit may comprise a reporter element characteristic of the identity of the respective nucleotide in the target nucleic acid. The plurality of subunits may comprise at least one selectively cleavable bond which may be cleaved to provide a polymer of an extended length, such as disclosed in WO2008157696. The polymer may be an expanded polynucleotide such as disclosed in WO2014096830, comprising in respect of each nucleotide in the target polynucleotide in sequence order, clock nucleotides and at least one signal nucleotide in a predetermined order, the clock nucleotides having a predetermined sequence common to each nucleotide in the target polynucleotide, and the at least one signal nucleotide that is characteristic of the identity of the respective nucleotide in the target polynucleotide. Methods to control translocation of the polymer may be employed such as disclosed in WO2015175789. The method may comprise use of a pulsing voltage. The polymer may comprise translocation control elements. In some examples, the method may include using a machine learning model that may be a classifier, and analysing the measurement signal may comprise classifying the measurement signal using the machine learning model to determine that the portion is associated with the first state. In such examples, a classification for portions of the measurement signal as being associated with respective nanopore states may obtained directly by processing the measurement signal using the machine learning model. In particular, a classifier based on a machine learning model may be trained to capture complex and non-linear relationships involve in determining the association of measurements with nanopore states. In this manner, the accuracy with which portions of the measurement signal are associated with states of the nanopore may be enhanced. The method may include analysing the measurement signal by processing the measurement signal by the machine learning model to determine values of a metric for measurements of the sample within the measurement signal, wherein determining that the portion of the measurement signal is associated with the first state of the nanopore may be based at least in part on the determined values of the metric. Determining values of a metric for the measurement signal using a machine learning model, such as a regressor, may enable mathematically complex metrics to be accurately determined for the measurement signal. In this manner, the method may be employed to enhance the accuracy of known metric-based methods for associating portions of the measurement signal with states of the nanopore. Such metric-based methods may be associated with physically meaningful measures enabling a user of the polymer analysis system to continue operating the system based on existing measures. According to another aspect of the present disclosure, there is provided a further computer-implemented method of processing a measurement signal comprising measurements of a sample by a sensor unit comprising a nanopore. There is also provided apparatus for carrying out the further method, and a computer program product (such as a non-transitory storage medium) comprising machine readable instructions which, when executed by a computer, cause the computer to carry out the further method. The further method includes processing the measurement signal to determine values of a metric, wherein the processing comprises at least one of: computing an autocorrelation of the measurement signal, or computing a number of crossings of a given value of the measurement signal. The method may include determining, using the determined values of the metric, that a portion of the measurement signal is associated with a first state of the nanopore, the first state being one of a plurality of states of the nanopore, and determining whether to further analyse the portion of the measurement signal in dependence on the determination that the portion of the measurement signal is associated with the first state of the nanopore. Computing an autocorrelation or a number of crossings of the measurement signal may be computationally less onerous compared with employing a machine learning model. In particular, using such techniques may help avoid the need to (and the computational resources required to) train a machine learning model. Moreover, such techniques may also involve vector operations that may, in some examples, have better time and space complexity compared with the techniques involved in the use of machine learning models relying on matrix operations. Further features and advantages of the invention will become apparent from the following description of preferred embodiments of the invention, given by way of example only, which is made with reference to the accompanying drawings. Brief Description of the Drawings Figure 1 shows a schematic of apparatus for analysing polymers. Figure 2 shows a flow diagram representing a first example method of operating a polymer analysis system. Figure 3 shows a schematic diagram of an example method of analysing polymers. Figure 4 shows a flow diagram representing a second example method of operating a polymer analysis system. Figure 5 shows a flow diagram representing operation decisions taken in dependence on processing a measurement signal obtained from a sensor unit. Figure 6A illustrates a first example of a measurement signal obtained from a sensor unit and a corresponding graph showing determined metric values. Figure 6B illustrates a second example of a measurement signal obtained from a sensor unit and a corresponding graph showing determined metric values. Detailed Description Details of systems and methods according to examples will become apparent from the following description with reference to the figures. In this description, for the purposes of explanation, numerous specific details of certain examples are set forth. Reference in the specification to ‘an example’ or similar language means that a feature, structure, or characteristic described in connection with the example is included in at least that one example but not necessarily in other examples. It should be further noted that certain examples are described schematically with certain features omitted and / or necessarily simplified for the ease of explanation and understanding of the concepts underlying the examples. Fig. 1 shows a schematic of apparatus 100 for analysing polymers. The system 100 comprises a flow cell 102 for taking measurements of a sample 104. In examples, the flow cell 102 may be a device that facilitates the flow of the sample 104 through one or more nanopores associated with sensor units for obtaining measurements of the sample 104. The sample 104 may be a fluid mixture, and may contain one or more sample components. Within the meaning of the present disclosure, a sample component may be a chemical substance such as a molecule, polymer, or compound, or any other constituent of the mixture. For example, the sample 104 may be an analyte of interest and may contain one or more polymers that a user may be seeking to analyse. Where the polymer of interest is a long chain of polymer units, it may be in fragmented form within the sample 104. In such examples, the apparatus 100 may be configured to analyse a composite polymer after assembling measurements of individual polymer fragments. Analysis of the polymers may involve sequencing the polymer units included in the one or more polymers. In such contexts, the flow cell 102 may be referred to as a sequencing flow cell. In examples, the sample 104 may include a polynucleotide (or nucleic acid), a polypeptide such as a protein, a polysaccharide, or any other polymer. The polymer may be natural or synthetic. The polymer units may be nucleotides. The nucleotides may be of different types that include different nucleobases. The polynucleotide may be deoxyribonucleic acid (DNA), ribonucleic acid (RNA), cDNA or a synthetic nucleic acid known in the art, such as peptide nucleic acid (PNA), glycerol nucleic acid (GNA), threose nucleic acid (TNA), locked nucleic acid (LNA) or other synthetic polymers with nucleotide side chains. The polynucleotide may be single-stranded, be double-stranded or comprise both single-stranded and double-stranded regions. Typically, cDNA, RNA, GNA, TNA or LNA are single stranded. In some examples, the sample 104 may include polymers other than polynucleotides, some non-limitative examples of which are as follows. The polymer may be a polypeptide, in which case the polymer units may be amino acids that are naturally occurring or synthetic. The polymer may be a polysaccharide, in which case the polymer units may be monosaccharides. Particularly where the apparatus 100 comprises a nanopore and a polymer in the sample 104 comprises a polynucleotide, the polynucleotide may be long, for example at least 5kB (kilo-bases), i.e. at least 5,000 nucleotides, or at least 30kB (kilo-bases), i.e. at least 30,000 nucleotides. As described above, the sequence of nucleotide units in DNA, for example, encodes information relating to the biological function of an organism. The apparatus 100 may enable sequencing of such biopolymers as DNA to support developments in the field of biotechnology. Beyond biological polymers, homopolymeric polysterene is brittle at room temperature but has a low glass-transition temperature at which it becomes soft and rubbery; copolymerised with acrylonitrile, however, the glass-transition temperature of polyseterene increases significantly, causing it to remain hard at higher temperatures. By doing so, polysterene can be used to produce containers for hot beverages. By enabling sequencing of synthetic polymers, the apparatus 100 may shed light on the dependence of their properties and their constitutive polymer units to support developments in materials technologies. In order to obtain measurements of the sample 104, the flow cell 102 may house or be connected to a sensor unit 106 that includes a nanopore 108. For example, the nanopore 108 may be set in an arrayed sensor chip. The nanopore 108 may be a pore that may have a size of the order of nanometres, allowing the translocation of sample components, such as polymers, therethrough. A property that depends on the molecular units of a sample component being translocated through the nanopore 108 may be measured. The property may be associated with an interaction between the component and the nanopore 108 during translation of the component through the nanopore 108. Interaction of the sample component may occur at a constricted region of the nanopore 108. For example, the sample may include a polymer that may be translocated through the nanopore 108. Translocation of the polymer through the nanopore 108 may be carried out using a method such as described in WO2021 / 111125, WO2021 / 133168 or WO2023 / 118891. By translocating the polymer through the nanopore 108, raw measurements may be obtained that are indicative of interactions between polymer units included in the polymer and the nanopore 108. A single raw measurement obtained from the sensor unit 106 may correspond to an individual polymer unit or a group of polymer units (from a neighbourhood) included in a polymer. In this manner, the polymer units may be read into a raw measurement signal by the sensor element of the sensor unit 106. The sensor unit 106 may transfer the measurement signal comprising raw measurements of a sample to a processing unit 110. The nanopore 108 may be situated in a membrane; for example, the nanopore 108 may be situated in a transmembrane nanopore. The nanopore 108 may function as an aperture in the membrane. For example, the nanopore 108 may be a protein pore. In such examples, the membrane may be flanked by ionic solutions and the membrane material may preferably have a high resistance or resistivity so that the material volume is inconducive to the flow of ions. As an example, the membrane may be constructed out of an amphiphilic layer, for example a lipid bilayer. An amphiphilic layer is a layer formed from amphiphilic molecules, such as phospholipids, which have both hydrophilic and lipophilic properties. The amphiphilic layer may be a monolayer or a bilayer. The amphiphilic layer may be a co-block polymer such as disclosed by (Gonzalez-Perez et al., Langmuir, 2009, 25, 10447-10450) or by PCT / GB2013 / 052767, published as WO2014 / 064444. Transmembrane protein pores for use in the methods described herein can be derived from P-barrel pores or a-helixbundle pores. P-barrel pores comprise a barrel or channel that is formed from P-strands. Suitable P-barrel pores include, but are not limited to, P-toxins, such as a-hemolysin (a-HL), anthrax toxin and leukocidins, and outer membrane proteins / porins of bacteria, such as Mycobacterium smegmatis porin (Msp), for example MspA, outer membrane porin F (OmpF), outer membrane porin G (OmpG), outer membrane phospholipase A and Neisseria autotransporter lipoprotein (NalP). a-helix bundle pores comprise a barrel or channel that is formed from a-helices. Suitable a-helix bundle pores include, but are not limited to, inner membrane proteins and a outer membrane proteins, such as WZA and Cly A toxin. The transmembrane pore may be derived from Msp or from a-HL. Any transmembrane pore may be used in the methods provided herein. The pore may be biological or artificial. Suitable nanopores include, but are not limited to, protein pores, polynucleotide pores, and pores formed in solid state substrates such as silicon nitride. A solid-state pore may comprise a nanochannel. The pore may be a DNA origami pore such as disclosed in WO2013 / 083983. The protein pore may be a monomer or an oligomer. The pore may be made up of several repeating subunits, such as at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, or at least 16 subunits. The pore may be a hexameric, heptameric, octameric, nonameric or decameric pore. The pore may be a homooligomer or a hetero-oligomer. The transmembrane protein pore may be modified from the wild type. The transmembrane protein pore may be derived from P-barrel pores or a-helix bundle pores such as a-hemolysin, anthrax toxin and leukocidins, and outer membrane proteins / porins of bacteria, such as Mycobacterium smegmatis porin (Msp), for example MspA, MspB, MspC, MspD, CsgG, outer membrane porin F (OmpF), outer membrane porin G (OmpG), outer membrane phospholipase A and Neisseria autotransporter lipoprotein (NalP) and other pores, such as lysenin, inner membrane proteins and a outer membrane proteins, such as WZA and ClyA toxin. The transmembrane protein pore may be derived from derived from Spl or haemolytic protein fragaceatoxin C (FraC). Examples of suitable pores derived from CsgG are disclosed in WO2016 / 034591, WO2017 / 149316, WO2017 / 149317, WO2017 / 149318 and WO2019 / 002893, each of which is hereby incorporated by reference in its entirety. The pore and / or enzyme may be modified to improve interaction between the two, such as disclosed in WO2015 / 166276. In further examples, an alternative type of pore may be used for obtaining measurements of the sample 104. A suitable transmembrane protein pore may be derived from Msp, preferably from MspA. Such a pore will be oligomeric and typically comprises 7, 8, 9 or 10 or more monomers derived from Msp. The pore may be a homo-oligomeric pore derived from Msp comprising identical monomers. Alternatively, the pore may be a hetero-oligomeric pore derived from Msp comprising at least one monomer that differs from the others. The pore may also comprise one or more constructs that comprise two or more covalently attached monomers derived from Msp. Suitable pores are disclosed in WO-2012 / 107778. The pore may be derived from MspA or a homolog or paralog thereof. The nanopore 108 may be a naturally occurring pore or may be a mutant pore. Typical pores are described in WO-2010 / 109197, Stoddart D et al., Proc Natl Acad Sci, 12; 106(19):7702-7, Stoddart D et al., Angew Chern Int Ed Engl. 2010;49(3):556-9, Stoddart D et al., Nano Lett. 2010 Sep 8; 10(9):3633-7, Butler TZ et al., Proc Natl Acad Sci 2008; 105(52):20647-52, and WO-2012 / 107778. The nanopore may be MS-(B1)8. The nucleotide sequence encoding Bl and the amino acid sequence ofBl are Seq ID: I and Seq ID: 2. The nanopore 108 is more preferably MS-(B2)8. The amino acid sequence of B2 is identical to that of BI except for the mutation L88N. The nucleotide sequence encoding B2 and the amino acid sequence ofB2 are Seq ID: 3 and Seq ID: 4. The nanopore 108 may also be a solid state pore comprising an aperture formed in a solid state layer. Such a solid state layer may not be of biological origin. In other words, a solid state layer may not be derived from or isolated from a biological environment such as an organism or cell, or a synthetically manufactured version of a biologically available structure. Solid state layers can be formed from both organic and inorganic materials including, but not limited to, microelectronic materials, insulating materials such as Si3N4, A1203, and SiO, organic and inorganic polymers such as polyamide, plastics such as Teflon® or elastomers such as two-component additioncure silicone rubber, and glasses. The solid state layer may be formed from graphene. Suitable graphene layers are disclosed in WO-2009 / 035647 and WO-2011 / 046706. When the solid state pore is an aperture in a solid state layer, the aperture may be modified, chemically, or otherwise, to enhance its properties as a nanopore. A solid state pore may be used in combination with additional components which provide an alternative or additional measurement of the polymer such as tunnelling electrodes (Ivanov AP et al., Nano Lett. 2011 Jan 12; 11(1):279-85), or a field effect transistor (FET) device (WO-2005 / 124888). Suitable solid state pores may be formed by known processes including for example those described in WO-00 / 79257. In the example of the apparatus 100 shown in Figure 1, the measurements may be electrical measurements, in particular current measurements of the ion current flowing through the nanopore 108. In general, these and other electrical measurements may be made using standard single channel recording equipment as describe in Stoddart D et al., Proc Natl Acad Sci, 12; 106(19):7702-7, Lieberman KR et al, J Am Chern Soc 2010; 132(50): 17961-72, and WO-2000 / 28312. Alternatively, electrical measurements may be made using a multi-channel system, for example as described in WO-2009 / 077734 and WO-2011 / 067559. Suitable conditions for measuring ionic currents through transmembrane protein pores are known in the art. The methods described below may be carried out with a voltage applied across the membrane and pore. The voltage used may vary from +2 V to -2 V, or from -400 mV to +400 mV. It is possible to increase discrimination between different polymer units such as nucleotides by a pore by using an increased applied potential. The methods described herein may be carried out in the presence of any charge carriers, such as metal salts, for example alkali metal salt, halide salts, for example chloride salts, such as alkali metal chloride salt. Charge carriers may include ionic liquids or organic salts, for example tetramethyl ammonium chloride, trimethylphenyl ammonium chloride, phenyltrimethyl ammonium chloride, or l-ethyl-3-methyl imidazolium chloride, potassium chloride, sodium chloride, caesium chloride or rubidium chloride. The salt concentration may be 3 M or lower and is typically from 0.1 to 2.5 M. Hel308, XPD, RecD and Tral helicases surprisingly work under high salt concentrations which advantageously provide a high signal to noise ratio. The methods may be carried out in the presence of a buffer. Any suitable buffer may be used in the method of the invention such as HEPES or Tris-HCl buffer. The pH may vary from 4.0 to 12.0 and is preferably about 7.5. The methods may be carried out at a temperature that supports enzyme function. The membrane may be supported on a support structure separating cis and trans compartments comprising ionic solution. An array of membranes each comprising a nanopore may be provided on a support structure comprising a common cis chamber and multiple trans chambers such as disclosed by WO2014 / 064443. The potential difference may be applied across the nanopore between electrodes provide in the cis and trans compartments. Suitable electrode materials include Pt, Pd and Ag. The electrodes may be reference electrodes such as Ag / AgCl. The ionic solution may comprise a redox couple such as potassium ferri / ferrocyanide. The nanopore 108 may be supported in a membrane having cis and trans openings on opposite sides of the membrane. The membrane may preferably be an amphiphilic layer. An amphiphilic layer is a layer formed from amphiphilic molecules, such as phospholipids, which have both hydrophilic and lipophilic properties. The amphiphilic molecules may be synthetic or naturally occurring. Non-naturally occurring amphiphiles and amphiphiles which form a monolayer are known in the art and include, for example, block copolymers such as disclosed in WO2014 / 064444. The block copolymer may be a diblock (consisting of two monomer sub-units), but may also be constructed from more than two monomer sub-units to form more complex arrangements that behave as amphiphiles. The copolymer may be a triblock, tetrablock or pentablock copolymer. The amphiphilic molecules may be chemically-modified or functionalised to facilitate coupling of a polymer such as a polynucleotide. The amphiphilic layer may be a monolayer or a bilayer. The amphiphilic layer may be planar. The amphiphilic layer may be curved. The membrane may be a lipid bilayer. The lipids may comprise a head group, an interfacial moiety and two hydrophobic tail groups which may be the same or different. Suitable lipid bilayers are disclosed in WO2008 / 102121, WO2009 / 077734 and WO2006 / 100484. Possible electrical measurements also include: nanopore tunnelling measurements, FET measurements and optical measurements combined with electrical measurements such as disclosed by WO2005 / 124888, WO2020 / 183172, Soni GV et al., Rev Sci Instrum. 2010 Jan;81(l):014301, Ivanov AP et al., Nano Lett. 2011 Jan 12;11(1):279-85 and WO2016 / 009180. As mentioned hereinbefore, the sample 104 may include a polymer that may be a polynucleotide. Polynucleotide as used herein refers to a polymeric form of nucleotides of any length, either ribonucleotides or deoxyribonucleotides. This term refers only to the primary structure of the molecule. Thus, this term includes double-and single-stranded DNA, and RNA. In some embodiments the polynucleotide is a single-stranded DNA-RNA hybrid. DNA-RNA hybrids can be prepared by ligating single-stranded DNA to RNA or vice versa. The polynucleotide may be doublestranded deoxyribonucleic acid (DNA) or double-stranded ribonucleic nucleic acid (RNA). In some embodiments the polynucleotide may be single-stranded DNA. In some embodiments the polynucleotide is double-stranded RNA. In some embodiments the polynucleotide may be a double-stranded DNA-RNA hybrid. Double-stranded DNA-RNA hybrids can be prepared from single-stranded RNA by reverse transcribing the cDNA complement. The polynucleotide can be of any length. For example, the polynucleotide can be at least 10, at least 50, at least 100, at least 150, at least 200, at least 250, at least 300, at least 400 or at least 500 nucleotides or nucleotide pairs in length. The polynucleotide can be 1000 or more nucleotides or nucleotide pairs, 5000 or more nucleotides or nucleotide pairs in length or 100000 or more nucleotides or nucleotide pairs in length. The polynucleotide may be made up of deoxyribonucleotide bases or ribonucleotide bases. Nucleic acids may be manufactured synthetically in vitro or isolated from natural sources. Nucleic acids may further include modified DNA or RNA, for example DNA or RNA that has been methylated, or RNA that has been subject to post-translational modification, for example 5’-capping with 7-methylguanosine, 3’-processing such as cleavage and polyadenylation, and splicing. Nucleic acids may also include synthetic nucleic acids (XNA), such as hexitol nucleic acid (HNA), cyclohexene nucleic acid (CeNA), threose nucleic acid (TNA), glycerol nucleic acid (GNA), locked nucleic acid (LNA) and peptide nucleic acid (PNA). Sizes of nucleic acids, also referred to herein as “polynucleotides” are typically expressed as the number of base pairs (bp) for double-stranded polynucleotides, or in the case of single-stranded polynucleotides as the number of nucleotides (nt). One thousand bp or nt equal a kilobase (kb). Polynucleotides of less than around 40 nucleotides in length are typically called “oligonucleotides” and may comprise primers for use in manipulation of DNA such as via polymerase chain reaction (PCR). Nucleotides can have any identity, and include, but are not limited to, adenosine monophosphate (AMP), guanosine monophosphate (GMP), thymidine monophosphate (TMP), uridine monophosphate (UMP), 5-methylcytidine monophosphate, 5-hydroxymethylcytidine monophosphate, cytidine monophosphate (CMP), cyclic adenosine monophosphate (cAMP), cyclic guanosine monophosphate (cGMP), deoxyadenosine monophosphate (dAMP), deoxyguanosine monophosphate (dGMP), deoxy thy mi dine monophosphate (dTMP), deoxyuridine monophosphate (dUMP), deoxycytidine monophosphate (dCMP) and deoxymethylcytidine monophosphate. The nucleotides are preferably selected from AMP, TMP, GMP, CMP, UMP, dAMP, dTMP, dGMP, dCMP and dUMP. A nucleotide may be abasic (i.e. lack a nucleobase). A nucleotide may also lack a nucleobase and a sugar (i.e. is a C3 spacer). The polynucleotide may be a concatemer comprising multiple sets of complementary strands each linked by a bridging molecule or moiety. The concatemer may be provided by methods such as disclosed by US9910956. The methods described herein may be used to identify or characterise any nucleotide included in a polynucleotide. The nucleotide can be naturally occurring or artificial. A nucleotide typically contains a nucleobase (which may be shortened herein to “base”), a sugar and at least one phosphate group. The nucleobase is typically heterocyclic. Suitable nucleobases include purines and pyrimidines and more specifically adenine, guanine, thymine, uracil and cytosine. The sugar is typically a pentose sugar. Suitable sugars include, but are not limited to, ribose and deoxyribose. The nucleotide is typically a ribonucleotide or deoxyribonucleotide. The nucleotide typically contains a monophosphate, diphosphate or triphosphate. The nucleotide can include a damaged or epigenetic base. The nucleotide can be labelled or modified to act as a marker with a distinct signal. This technique can be used to identify the absence of a base, for example, an abasic unit or spacer in the polynucleotide. Of particular use when considering measurements of modified or damaged DNA (or similar systems) are the methods where complementary data are considered. The additional information provided allows distinction between a larger number of underlying states. The system 100 may include electrical circuitry in the sensor unit 106 to control translocation of sample components through the nanopore 108. For example, polynucleotides such as DNA or RNA included in the sample may have an associated electric charge, e.g. a negative charge, and this property may be exploited to control translocation of the polynucleotide by adjusting a potential difference across the nanopore. The electrical circuitry may include electrodes for providing a potential difference across the nanopore 108 to translocate the polynucleotide through the nanopore 108. For example, the system 100 may be operated by controlling a bias control circuit in the sensor unit 106 to apply a bias voltage across the nanopore 108 that may be sufficient to cause translocation of the polynucleotide. Likewise, the potential difference applied across the nanopore 108 may be reversed during translocation of the polynucleotide, causing it to be ejected from the nanopore 108. For example, the bias control circuit may be controlled to apply a bias voltage across the nanopore pore 110 that is sufficient to eject the polynucleotide currently being translocated. By ejecting the polynucleotide occupying the nanopore 108, the nanopore 108 may become available to receive a further polynucleotide from the sample 104. For example, after ejection, the bias control voltage (applied potential difference) may be reset to a value that favours or drives translocation of a further polynucleotide from the sample 104 through the nanopore 108. Thus, a polymer in the nanopore 108 may be ejected and re-translocated through the nanopore 108. Similar methods may be used for translocating polymers other than polynucleotides through the nanopore 108. In some examples, the nanopore 108 may further facilitate translocation under the control of enzymatic activity, such as under the influence of a motor enzyme, as described hereinbefore. A polynucleotide binding protein such as helicase may be provided in the nanopore 108 to provide for movement of a polynucleotide through the nanopore 108 in a stepwise fashion. As an example, a combination of applied potential difference and enzymatic activity may be used to carry out translocation of a polynucleotide included in the sample 104. In some examples, a detection channel may be included in the nanopore sensor unit 108 which may detect translocation so that measurements can start to be taken. Similar methods may be used for controlling and detecting translocation of non-polynucleotide polymers through the nanopore 108. Suitable polynucleotide binding proteins for guiding translocation of the polynucleotide are known in the art and include polymerases, exonucleases, helicases and topoisomerases. A preferred enzyme is a helicase which may be or be derived from a Hel308 helicase, a RecD helicase, such as Tral helicase or a TrwC helicase, a XPD helicase or a Dda helicase. The helicase may be any of the helicases, modified helicases or helicase constructs disclosed in WO2013 / 057495, WO2013 / 098562, WO2013 / 098561, WO2014 / 013259; WO2014 / 013262 and WO2014 / 013260. The helicase may be added to the polymer during sample preparation and stalled by one or more spacers as disclosed in WO2014 / 135838. Motor enzymes other than helicase may alternatively be used. Translocation of the polymer may be guided by alternative methods disclosed for example in WO2019 / 006214 and WO2020 / 016573. All publications, patents and patent applications cited herein, whether supra or infra, are hereby incorporated by reference in their entirety. The electrical circuitry in the sensor unit 106 may include a sensor element to measure a current flowing through the nanopore 108 during translocation sample components. For example, a current flowing in a transverse direction of the nanopore 108 or in a surrounding region of the nanopore 108 may be measured during translocation of a polymer. The current may vary in response to the passage of each polymer unit of the polymer within the nanopore 108, and may do so to an extent that corresponds with a particular polymer unit or multiple units being translocated through the nanopore 108. For example, variations of current may be on the order of picoamperes, nanoamperes, or microamperes. The ionic current may be measured by the sensor, and the resulting measurement signal may optionally be amplified using an amplifier, such as a low-noise amplifier system. In alternative examples, the measurable property may be a voltage difference or a change in resistivity, conductance or other ionic property of the nanopore 108 or a suitable ionic property of the surrounding medium of the nanopore 108. Variations in the measurement property of the nanopore 108 may be indicative of one or more different states of the nanopore 108. For example, certain values (or certain ranges of values) of the current across the nanopore 108 may be typical of a polynucleotide being translocated through the nanopore 108. A sequence of measurements exhibiting such values may serve as an identifiable fingerprint for a state of the nanopore 108 representative of the polynucleotide being translocated through the nanopore 108. Similarly, other states of the nanopore 108, including states the pore as being blocked, may be identifiable from the measurement signal and variations of measured values therein. After optionally amplifying the measurement signal, the signal may be digitised and provided to a processing unit 110 for analysis (indicated using an arrow from the sensor unit 106 to the data processing system 112 in Figure 1). For example, the sensor unit 110 may include an Application Specific Integrated Circuit (ASIC) for obtaining measurements. The processing unit 110 may comprise a data processing system 112 containing hardware and software components capable of processing the measurement signal obtained from the sensor unit 106 (or ASIC thereof). The data processing system 112 may be a desktop, a laptop, a tablet, or a server, or any combination of the above. In some examples, the data processing system may be a dedicated device for sequencing polymers, such as produced by Oxford Nanopore Technologies (ONT). The data processing system 114 may include a central processing unit (CPU) having one or more processors including integrated circuit microprocessors that may optionally be multithreaded, such as Intel Xeon or Intel Core i3 / i5 / i7 series processors. The data processing system 116 may also optionally include one or more graphic processing units (GPUs) such as Nvidia A100 or H100 or H800 GPUs or AMD GPUs. Similarly, the data processing system 114 may contain software that may include source code, object code, firmware, etc. in any suitable language. For example, the source code may be written in Python, C, C++, Rust, Julia, etc and may use specific development frameworks, libraries, or packages, including PyTorch, TensorFlow, Keras, CUDA, etc. The examples provided herein for the hardware and software components do not constitute an exhaustive list; many more similar components may be alternatively or additionally included. The data processing system 112 may include stored software components such as machine learning models or algorithms. Such software components may be capable of sequencing polymers included in the sample 104 by processing measurements of a polymer being translocated through the nanopore 108. For example, the data processing system 112 may be capable of basecalling polynucleotides such as DNA by analysing relevant portions of the measurement signal. The software components may also be capable of assembling measurements obtained from multiple sample components, such as polymer fragments of a composite polymer, prior to sequencing or classifying the composite molecule. The data processing system 112 may be capable of classifying components of the sample 104. For example, the data processing system 112 may be capable of classifying peptides and polypeptides. In these examples, the data processing system 112 may be additionally capable of pre-processing the measurement signal obtained from the sensor unit 106. In particular, algorithms programmed into the data processing system 112 may annotate portions of the measurement signal. For example, some portions of the measurement signal may be flagged as being relevant for basecalling and classification, whereas others may be annotated as being irrelevant and worthy of omission. In this manner, the basecaller or classifier may dedicate its resources to computational processing of portions of the measurement signal containing measurements of a translocating polymer. Furthermore, the data processing system 112 may be capable of instructing the sensor unit 106, in response to the annotations made available from the pre-processing, to change the state of the nanopore 108. For example, the sensory unit may be capable of instructing the nanopore 108 to cause sample components to be ejected from the nanopore 108. In this manner, the nanopore 108 may dedicate its operation time to translocation of polymers rather than in jammed states. Pre-processing conducted by the data processing system 112 may therefore improve the overall sequencing or classification yield of the flow cell 102. Methods described herein are directed to improving the yield by implementing appropriate preprocessing algorithms and responsive instruction protocols in the apparatus 100. A method 200 of operating a polymer analysis system 100 will now be described with reference to Fig. 2. The method 200 involves obtaining, at 214, a measurement signal comprising measurements of the sample 104 by the sensor unit 106. The measurements may include raw measurements indicative of interactions between one or more sample components and the nanopore 108 included in the sensor unit 106. As an example, the measurement signal may include values of the ionic current as measured across the nanopore 108 over a certain period of time. In examples, the measurement signal may be a continuous time signal spanning a given period of operation of the sensor unit 106. In other examples, the measurement signal may be assembled from signal segments corresponding to different measurement periods. The measurement signal may also be a partition of a longer duration signal; such partitioning may be performed, for example, to facilitate parallel processing of an original measurement signal. The measurement signal may have been processed to make it suitable for analysis. For example, the measurement signal may have been processed to filter out noise and / or other undesirable numerical artefacts. In examples, the measurement signal may include measurements of a polymer in the sample 104. For example, the sample 104 may contain a polynucleotide that needs to be sequenced. The polynucleotide may be present in the sample 104 in addition to various other sampler components, including polymers as well as nonpolymers. Different portions of the measurement signal may thus correspond to different states of the nanopore 108 representing interactions between the nanopore 108 and respective sample components. For example, in some portions, the nanopore 108 may exhibit a state corresponding to an interaction between the nanopore 108 and the polynucleotide. In such examples, the state of the nanopore 108 may be representative of the polynucleotide being translocated through the nanopore 108. Sequencing the polynucleotide may require processing portions of the measurement signal corresponding to such a state. Similarly, in other portions of the measurement signal, the nanopore 108 may exhibit a state that is representative of interactions between the nanopore 108 and other components of the sample 104. It is noted that the other sample components may have been introduced intentionally during sample preparation. For example, some molecules may be necessary for maintaining the thermo-chemical stability of the sample 104 or polymers therein. Alternatively, components may be introduced in the sample 104 as inevitable by-products of sample preparation. In any case, the measurement signal may include measurements of the polymer (e.g. a polynucleotide) as well as measurements other sample components. Regardless of how the various sample components are introduced in the sample 104, the measurement signal obtained from measuring the sample 104 may include measurements corresponding to various states of the nanopore 108. Depending on the objective of an analysis task, it may be desirable to process the measurement signal to identify measurements corresponding to the relevant states of the nanopore 108. Fig. 3 shows a schematic diagram of an example method 300 of analysing polymers. The flow cell 302 is shown containing a sample 304, which includes a polymer 316. The nanopore 308 is shown to exhibit three different states 318, 320 and 322 over a period of operation of the flow cell 302. As mentioned above, each nanopore state corresponds to a respective class of interaction between the nanopore 308 and the sample 304. For example, 318 is a nanopore state representative of the polymer 316 being translocated through the nanopore 308. In some examples, the polymer 316 may be a strand of DNA, which in state 318, may be in the process of translocating through the nanopore 308. Measurements obtained from the sensor unit 106 corresponding to state 318 may thus contain measurements of the sequence of nucleotides included in the polynucleotide 316. Thus, it may be profitable to process measurements corresponding to such a STRAND state 318 in order to sequence the polynucleotide 316. Likewise, it may not be particularly profitable to process measurements corresponding to other states that are not representative of the DNA strand being translocated through the nanopore 308. In order to improve the sequencing efficiency, it may therefore be desirable to process solely the portions of the measurement signal that correspond to a STRAND state of the nanopore 308. Where the polymer 316 is not a polynucleotide, states representative of a translocating polymer may be referred to as a POLYMER or a TRANSLOCATING or, simply, a USEFUL state. The nanopore state 320 is representative of the nanopore 308 wot being occupied by the polymer 316. In such an OPEN state, the polymer 316 may not interact measurably with the nanopore 308. Portions of the measurement signal associated with the OPEN state may therefore include no measurements of polymer units included in the polymer 316. Instead, such portions may include measurements of components other than the polymer 316 present in the sample. In some examples, the polymer 316 may be appended with molecules during sample preparation to assist translocation of the polymer 316 through the nanopore 108. Translocation of such appended molecular components may also correspond with either a NON-STRAND state of the nanopore 108. For example, a DNA polynucleotide may be appended with an adapter for linking one DNA strand with another. In other examples, an adapter may be appended to a polynucleotide strand to enhance the ability of the strand to be translocated through the nanopore 108. Such adapter molecules may themselves contain nucleotides but may not, in general, reflect the nucleotide sequence of the polynucleotide. Where sequencing of the polynucleotide is intended, it may be desirable to omit processing the portions of the measurement signal associated with an OPEN or a NON-STRAND state. For example, it may be advantageous to filter out such portions of the measurement signal prior to further downstream processing directed to sequencing. In other examples, it may be preferable to annotate the signal with states associated with each measurement or signal portion, while maintaining the length of the signal. Fig. 3 shows state 322 as being representative of the polymer 316 being blocked in the nanopore 308. In such BLOCKED states, the polymer 316 may be stuck within the nanopore 308 in a manner that prevents it to be translocated therethrough. The polymer units of the polymer 316 may interact with the nanopore 308, and may therefore be measured during a BLOCKED state. However, the polymer units may be measured in a fashion that is inconsistent with the sequence in which they appear in the polymer 316. Thus, it may be desirable to omit measurements corresponding to such BLOCKED states from processing directed to basecalling. In some examples, the polymer 316 may be blocked temporarily, such that the polymer 316 may be un-blocked through appropriate variations of the potential difference across the nanopore 308. In other examples, a polymer 316 may be jammed inside the nanopore 308 in a manner that permanently prevents further translocation. In such examples, it may be desirable to instruct the sensor unit of the flow cell 302 to completely eject the polymer 316. Thus, it may be advantageous to process portions of the measurement signal associated with BLOCKED states in order to carry out operations directed to resolving the BLOCKED state, subsequently allowing the nanopore 308 to translocate another sample component, such as the polymer 316. Fig. 3 shows an illustrative measurement signal 324 obtained during a period of operation of the flow cell 302. The signal 324 contains measurements of the sample 304, which may correspond to different states acquired by the nanopore 308 during the period of operation of the flow cell 302. A sensor unit including the nanopore 308 may measure changes in the ionic current across the nanopore 308 in response to one or more constitutive units of the component moving through the nanopore 308. For example, the value of ionic current may change in response to one or a group of polymer units being translocated through the nanopore 308. For a polynucleotide, each measurement in the measurement signal 324 may correspond to anywhere between 1 and 20 nucleotide bases from a neighbourhood of the polynucleotide. These numbers may vary in dependence on the chemical constitution of the polymer 316 being translocated, the rate of translocation, and various other physico-chemical parameters linked to the nanopore 308 and the polymer 316. Measurements included in the measurement signal 324 may be indicative of the state of the nanopore 308. For example, measurements of nucleotide bases included in a polynucleotide may typically exhibit values from a range of ionic current values. As an example, an ionic current value of less than 50 picoamperes may be indicative of the measurement corresponding to one or more nucleotide bases. A significantly different value of the ionic current may be indicative of the measurement corresponding to a different sample component. In other examples, the nanopore 108 may exhibit a STRAND state for a value of the ionic current that is higher than 50 picoamperes. Such differences may arise due to chemical or structural differences between two nanopores, e.g. due to differences in the membrane employed, but may also arise as a function of time during the course of operation of a given nanopore. In some contexts, the criteria for determining whether a measurement corresponds to a given state may be dependent on multiple parameters. Nevertheless, it may be useful to associate measurements included in the measurement signal 324 with respective states of the nanopore 108, e.g. by grouping measurements into “chunks” and annotating each chunk with a corresponding state of the nanopore 308. Chunking portions of the measurement signal 324 that correspond to a STRAND state may enable computational resources for sequencing to be dedicated to the measurements included in such portions. Fig. 2 shows that the method 200 includes, at 226, analysing the measurement signal using a machine learning model. The analysis may help determine that a portion of the measurement signal is associated with a state of the nanopore 108, the state being one of a plurality of states of the nanopore 108. For example, the analysis may result in outputs that help determine that a portion of the measurement signal is associated with a STRAND state or an OPEN state or a BLOCKED state of the nanopore 108. In other examples, the portion may be associated with other states. As will be described in detail below, outputs of the machine learning model may either directly associate portions of the measurement signal 324 with corresponding nanopore states, or may signify properties of the measurements that may indirectly help determine an association between the portions and corresponding nanopore states. Fig. 3 shows the measurement signal 324 being processed by a machine learning (ML) model 330. In one example illustrated within Fig. 3, the machine learning model 330 processes the measurement signal 324 and directly produces one or more classifications (indicated by the solid arrow from 330 to 332). More precisely, the machine learning model 330 may be a classifier that assigns classifications to one or more portions of the measurement signal 324. In such examples, the classification module 332 may serve to receive and transfer the classification outputs of the machine learning model 330. Each classification output may signify that a given measurement or group of measurements is associated with a given state of the nanopore 308. As an example, the machine learning model 330 may assign a STRAND classification to one or more portions of the measurement signal 324 that have been determined as being representative of the polymer 316 being translocated. Likewise, other portions of the measurement signal 330 may be assigned with an OPEN or a BLOCKED classification depending on whether the corresponding measurements are associated with an OPEN or a BLOCKED nanopore state. The machine learning model 330 may be capable of annotating the entire measurement signal 324 in this manner, assigning to each portion of the measurement signal 324 a respective classification that associates the portion with one of a plurality of states of the nanopore 308. In particular, the machine learning model 330 may be capable of processing a continuous time measurement signal 324 being obtained on-the-fly from the sensor unit. The annotations may be factored into further analysis, such as by sequencing algorithms that process the measurement signal 324 further. In other examples, the classification module 332 may transfer only the portions of the measurement signal 324 that are associated with the STRAND state to the sequencing algorithms. In this manner, the available computational resources may be dedicated to the portions of the measurement signal 324 containing measurements of the polymer 316 during translocation. In the examples described above, the machine learning model 330 may include one or more classification algorithms. In some examples, a binary classifier may be included. In such examples, the machine learning model 330 may associated each portion of the measurement signal 324 with either a STRAND or a NON-STRAND state of the nanopore 308. In other examples, a multi-class classifier may be adopted, which may assign fine-grained classifications to each portion of the measurement signal 324. As an example, the machine learning model 330 may determine that a portion of the measurement signal 324 is associated one of further sub-classes of the STRAND state of the nanopore 308. For example, a “High-complexity” state may be associated with a polynucleotide with a complex sequence of nucleotide bases being translocated through the nanopore 308. Likewise, a “Low-complexity” state may be associated with a polynucleotide with a simple sequence of nucleotide bases, such as a sequence including only two of the five canonical nucleobases. Similarly, fine-grained classifications may be assigned in contexts where the polymer 316 is not a polynucleotide. Such fine-grained classifications may be valuable for downstream processing decisions. For example, portions of the measurement signal 324 classified as low- or high-complexity state may be provided to different basecallers that are specifically trained to classify low- or high-complexity polynucleotides. In this manner, classification of portions of the measurement signal 324 may allow downstream processing and / or basecalling to be optimised. Processing the measurement signal 324 using a machine learning model 330 may, in some cases, reduce the computational expense associated with using hard-coded metrics to determine associations of the measurements and respective nanopore states. This is because, associating measurements based on hard-coded metrics may require measurements to be pre-processed to chunk the measurement signal 324 into “events”. Such events may be identified based on complex mathematical models that determine whether a given portion of the measurement signal 324 is associated with a given category of interaction between the nanopore 308 and the sample. For example, a sequence of up to 10 measurements may be chunked into an event that relates to the measurement of a polymer unit, or a non-polymer molecule. Processing the measurement signal to determine events for applying hard-coded metrics may be computationally expensive. It may instead be more expedient to operate directly on raw measurements, without clustering the signal into events, using a machine learning model. In this manner, the resource efficiency of the analysis system may be higher when using a machine learning model rather than a hard-coded metric. Moreover, a classifier based on a machine learning model 330 may be capable of accurately assigning classifications to portions of the measurement signal 324 compared with the use of hard-coded metrics. For example, according to a widely applied hard-coded metric, portions of the measurement signal 324 with ionic current values lying within a fixed range of values may be determined to be associated with a BLOCKED state of the nanopore 308. However, such a hard-coded metric may prove to be inaccurate when applied rigidly to all contexts. For example, the value of ionic current may drift over a period of operation of the nanopore 308 due to various temporal changes in the chemistry of the analyte. In some cases, the current across the nanopore 308 may drift due to depletion of a mediator complex in the solution. As a further example, the current passing through the nanopore 308 may drift during the course of operation of the flow cell 302 due to depletion of the redox chemistry in the bulk solution. In such examples, the values of measurements corresponding to a polynucleotide being translocated through the nanopore 308 may change over a period of time. Measurements corresponding to the STRAND state may be correctly classified at early stages of a run based on the hard-coded criterion, whereas at later stages, classification may be erroneous due to the overall drift in current values with time. The machine learning model 330 may be capable of accounting for such trends in the measurement signal 324. In examples, the measurement signal 324 may be associated with timestamps indicating times that respective measurements were taken by the sensor unit of the flow cell 302. In such examples, analysing the measurement signal 324 may comprise processing the timestamps in combination with the measurement signal 324 using the machine learning model 330. Such a machine learning model 330 may be trained a priori to classify portions of the measurement signal 324 by processing the measurements together with a respective timestamp. For example, a machine learning model 330 including a neural network may be trained using supervised learning methods. In such examples, a training signal may be provided together with annotations indicating training labels for measurements included in the signal, and the classification algorithm may be trained by optimising model parameters using backpropagation of loss function values computed based on predicted outputs. In this manner, the machine learning model 330 may account for temporal changes in the measurement signal 324 that may otherwise disrupt the performance of hard-coded metrics for classifying measurements. Furthermore, even disregarding intra flow cell variations in the current passing through a nanopore channel, there may be variations between different channels or flow cells, such as in cases where multiple flow cells are used to measure the polymers. In such examples, the baseline values for hard-coded metrics may vary due to flow cell to flow cell variations in nanopore sensitivities and membrane parameters. In such contexts, the use of hard-coded metrics may be infeasible or at least challenging. A machine learning model 330 may be able to address this problem by capturing such inter flow cell variations. In any case, whether there are intra flow cell or inter flow cell variations, the machine learning model 330 may be able to standardise a measurement signal based on accurately identifying the trends and / or variations and discounting them from the measurement signal. For example, the machine learning model 330 may process the measurement signal to identify a drift in the current passing through the nanopore and may subsequently correct the measurement signal to eliminate the drift. Similarly, the machine learning model 330 may process multiple measurement signals each obtained from a respective different flow cell to identify any inter flow cell variations and subsequently alleviate the signals of such variations. For example, a mux scan may be performed to identify differences in baseline metrics for different channels or flow cells. In some examples, a membrane-only channel or flow cell may be used to obtain baseline metrics. The variations arising from differences in such baseline metrics between different flow cells may then be identified and remove from the signal using the machine learning model 330. In this manner, a “recalibrated” measurement signal may be obtained, enabling hard-coded metrics to be applied without compromising the accuracy of classification / read splitting. In the examples described thus far, the machine learning model 330 may include one or more standard classification algorithms. For example, the machine learning model 330 may rely on logistic regression, k-nearest neighbours (KNN) or support vector machines (SVM). Alternatively, the machine learning model 330 may include neural networks that output the classification based on weights learned during training. Regardless of the classification algorithm adopted, the machine learning model 330 may be capable of accurately associating measurements with the corresponding nanopore states based on categorical mappings learned during training. As mentioned above, the association of portions or chunks of the measurement signal with respective states of the nanopore may support downstream processing decisions. Fig. 2 shows that the method 200 involves determining, at 234, whether to further analyse the portion of the measurement signal. In particular, whether to further analyse the portion may be determined in dependence on the determination that the portion of the measurement signal is associated with the state of the nanopore. For example, the method 200 may include determining to further analyse the portion in dependence on determining that the portion is associated with a state representative of a polymer being translocated through the nanopore 108. In the specific use case of polymer sequencing, further analysing the portion may include basecalling the portion of the measurement signal 324 to estimate a sequence of polymer units of the polymer 316. As an example, the method 200 may determine to further analyse one or more portions associated with the STRAND state when the polymer included in the sample is a polynucleotide. In other examples, the method 200 may determine not to further analyse the portion where the portion has been determined not to be associated with a STRAND state. Such decisions are optional to the method 200; further examples of downstream analysis decisions are described in detail below. Regardless, by limiting further analysis to portions associated with the STRAND state, the method may focus processing on measurements obtained during translocation of the polymer. In this manner, the sequencing yield of the flow cell 302 may be enhanced. Fig. 3 shows a basecaller 336 being provided with the outputs of the classification module 332. Where the machine learning model 330 is a classifier as described thus far, the classification module 332 may transfer one or more portions of the measurement signal 324 determined as being associated with a state representative of the polymer 316 being translocated through the nanopore 308. In particular, the classification module 332 may transfer such portions to the basecaller 336 for sequencing. The basecaller 336 may include a second machine learning model, which processes the transferred portion of the measurement signal 324 to determine a read sequence 338. In the example shown in Fig. 3, the read sequence 338 is a sequence of nucleotide bases obtained by processing portions of the measurement signal 324 that were classified by the machine learning model 330 as being associated with the STRAND state. In other examples, a shorter or longer read sequence may be obtained depending on the portions classified as being associated with the STRAND state. In relatively less frequent cases, the entire measurement signal 324 may correspond to a DNA strand being translocated through the nanopore 308. In such cases, the read sequence 338 obtained by relying on the classification module 332 may be identical to a read sequence obtained by feeding the measurement signal 324 directly to the basecaller 336. However, in most cases, the measurement signal 324 may correspond to a considerably long period of operation of the flow cell 302, and only parts of such periods may be associated with the polymer 316 being translocated through the nanopore 308. Not uncommonly, none of the measurements in the measurement signal 324 may be associated with the polymer 316 being translocated through the nanopore 308. In such cases, where a fraction of the measurement signal 324 is associated with states representative of the polymer 316 not being translocated through the nanopore 308, relying on the classification module 332 may be expected to reduce the computational processing by effectively filtering out irrelevant portions of the measurement signal 324. In this manner, the method 300 may improve the efficiency of sequencing and increase the sequencing yield of the flow cell 302. In particular, the basecaller 336 need not be trained to be able to perform the processing conducted by the classifer that assigns the classifications. Therefore, by producing classifications for the measurement signal 324 and / or filtering out irrelevant portions of the measurement signal 324, the machine learning model 330 enables the use of a lighter basecaller 336. As a result, training and optimisation of the basecaller 336 can focus on sequencing accuracy rather than on classifying nanopore states. Furthermore, by accurately classifying portions of the measurement signal 324 as being associated with the STRAND state using a machine learning model 330, the method 330 may predict a sequencing read 338 with higher fidelity. The passages above describe the use of a classifier machine learning model 330 for analysing the measurement signal and directly associating portions of the measurement signal 324 with a state of the nanopore 308. More generally, however, the machine learning model 330 may play a limited and / or supporting role in associating portions of the measurement signal 324 with nanopore states. For example, the machine learning model 330 may be a regressor. In such examples, the machine learning model 330 may process the measurement signal 324 to provide values of informative metrics that may indirectly aid classification of different portions of the measurement signal 324. In particular, Fig. 3 shows the measurement signal 324 being processed by the machine learning model 330 to compute values of a metric provided to a metric module 340. Values of the metric may be computed for measurements of the sample 304 within the measurement signal 324, and determining that the given portion of the measurement signal 324 is associated with a state of the nanopore 308 may be based at least in part on the computed values of the metric. The experimentally observed phenomenon of current drift is recalled here in the context of computing metric values. As explained above, measurements of current obtained from a nanopore 308 may exhibit an overall drift during the course of operation of the flow cell 302 comprising the nanopore 308. A hard-coded metric based on current values lying within a range may therefore fail to correctly predict the STRAND state classification for measurements throughout the period of operation. This is because the hard-coded metric may rely on a fixed range that does not account for drifts and other complex trends overlaid on the measured values. However, by a regressor machine learning model 330 may be capable of accounting for non-linear trends when processing the measurement signal 324 to more accurately derive the classification for portions of the measurement signal 324. In particular, the regressor machine learning model 330 may determine that current variations lying within a range that is merely offset from the hard-coded range due to an overall current drift correspond to measurements of a state representing the polymer 316 being translocated through the nanopore 308. In this manner, the machine learning model 330 may provide accurate computations of known metrics that indirectly enable association of the measurement signal 324 with one or more nanopore states. In such examples, the metric module 340 may transfer computed values of the metric to the classification module 332, which may then be able to derive classifications for each portion of the measurement signal 324 based on one or more threshold values for the metric. Training and using a regressor may be computationally less resource-intensive than training a classifier, thereby improving the efficiency of the system further. Moreover, the use of the machine learning model 330 as a regressor may enable existing metrics to be used (albeit more accurately) for determining associations between measurements or signal portions and the states of the nanopore 308. In this manner, existing analysis frameworks can be harnessed for the polymer analysis. In cases where polynucleotide sequencing is involved, adapting an existing framework may provide to be resource intensive and may create new sources of computational inefficiencies. By allowing a machine learning model 330 to be used to improve the accuracy of an existing framework, the method 300 may mitigate such inefficiencies. In examples, a regressor machine learning model 330 may be used to computed other relevant metrics for associating portions of the measurement signal 324 with respective nanopore states. For example, the machine learning model 330 may be trained to compute a dwell time of a polymer 316 included in the sample 304. In the context of nanopore sequencing, the dwell time of a molecule may be defined as the time a molecule spends within the nanopore during a sequencing run. The dwell time may be an important parameter that can be used to enhance the sequencing accuracy for a given set of measurements. For example, DNA basecallers may be trained to include parameters such as dwell time in the estimation of polynucleotide sequences. By training the machine learning model 330 to accurately calculate the dwell time, the method 300 may be able to rely on such parameters for sequencing. In this manner, the accuracy of downstream analysis (e.g. polynucleotide basecalling) may be enhanced. In further examples, the metric values computed by the machine learning model 330 may be indicative of a confidence measure. For example, the machine learning model 330 may be trained to regress to a confidence in a given measurement being associated with the polymer 316. By processing the measurement signal 324 using such a machine learning model 330, confidence values for each portion of the measurement signal 324 may be calculated. The calculated confidence values may be informative for further downstream tasks, such as basecalling. For example, a basecaller may be trained to rely on confidence values for a given set of measurements to estimate a corresponding sequence of nucleotide units. In some examples, the metric value calculated for each measurement in the measurement signal 324 may be a vector indicating multiple quantities. For example, the metric value may comprise a series of confidence values representing a confidence in the measurement being representative of each of a plurality of states of the nanopore 308. In other examples, each confidence value may represent a confidence in the corresponding measurement representing a particular class of polymers. For example, a polymer 316 may be composed of high- or low-complexity regions. Such regions may therefore correspond to polymer classes of different complexities. High-complexity polymers may involve a relatively higher number of different polymer units that repeat in a manner that is difficult to predict. Low- complexity may involve relatively fewer polymer units but may be challenging to distinguish from other states of the nanopore 308. For example, measurements of low-complexity regions of a polynucleotide being translocated through the nanopore 308 may be challenging to distinguish from measurements of the polynucleotide being blocked in the nanopore 308. In such examples, the confidence value associated with a measurement may represent a confidence in the measurement corresponding to a high-complexity or a low-complexity polymer region. In such examples, the metric may be a confidence value, but may also include a plurality of confidence values each indicating a confidence in the corresponding measurement being associated with a respective polymer complexity class. The metric may provide information for downstream processing tasks, such as basecalling. As an example, downstream processing of measurements associated with the polymer 316 may be configured based on the type of polymer region identified for the measurements. Thus, downstream processing tasks such as basecalling may be tailored to the complexity of the polymer regions being sequenced. In this manner, computation of metric values using the machine learning model 330 may enable the method 400 to be customised based on polymer complexity. For example, the method may rely on using less resource intensive basecallers for low-complexity regions and more accurate basecallers for high-complexity regions. As a result, the method 400 may be able to balance computational expense with accuracy. In alternative examples, polymer classes may be defined based on properties other than complexity. For example, different classes may correspond to different types of repeating polymer units included in the polymer 316. In one case, polymer classes may help distinguish between polynucleotides and polypeptides. The machine learning model 330 may be trained to compute confidence values for measurements being representative of one polymer class or the other. In such cases, a high confidence in a region being associated with a polynucleotide or a polypeptide may be relied upon for the purpose of basecalling. Thus, either by serving as a classifier or by serving as a regressor, the machine learning model 330 may be capable of pre-processing the measurement signal 324 to support downstream processing tasks such as classification. In particular, the machine learning model 330 may be facilitate association of chunks of measurements in the measurement signal 324 to be associated with respective states of the nanopore 308. In this manner, the machine learning model 330 may enable splitting of the measurement signal (i.e. perform “read splitting”) containing raw measurements obtained from the sensor unit into portions associated with respective nanopore states. Subsequent processing may then be performed on the basis of the read splitting obtained using the machine learning model 330. By splitting the read accurately, processing may be limited to portions of the measurement signal 324 that are associated with the polymer 316 being translocated through the nanopore 308. Improved read splitting may help better distinguish data corresponding to polymer translocation from “junk” data (corresponding to irrelevant sample components), so that inadvertent dropouts in coverage of polymer sequences may be reduced. In this manner, the sequencing efficiency and accuracy of the flow cell 302 may be enhanced considerably. It is noted that the method 200 may use further metrics to supplement those computed using the machine learning model 330. For example, the machine learning model 330 may associate a given portion of the measurement signal 324 with a state of the nanopore 308 representative of the polymer 316 being translocated. A length of the portion of the measurement signal may be computed. The portion length may serve as a metric indicative of whether or not the portion corresponds to an adapter being translocated through the nanopore 308. For example, the adapters appended to polymers 316 included in a given sample 304 may be known to have a fixed value of length (or, equivalently, include a maximum number of polymer units). Portions of the measurement signal 324 with lengths (or polymer unit count) close to the fixed value may be considered to be associated with nanopore states representative of the adapter rather than the polymer units of the polymer 316 being translocated through the nanopore 308. Subsequently, the method 200 may involve determining not to further analyse the portion of the measurement signal when the length of the portion matches an expected length for the adapter. Such techniques may be used in conjunction with the associations determined by the machine learning model 330 to enhance the overall accuracy of analysing the measurement signal 324. Whilst the passages above discuss the use of a machine learning model for the pre-processing of a measurement signal, the passages below discuss alternative means by which improved read splitting may be achieved. Fig. 4 shows a flow diagram representing a second example method of operating a polymer analysis system. In particular, this example supplants the use of a machine learning model 330 with specific metrics that, nevertheless, also circumvent the challenges of using hard-coded metrics as described previously. Similar to the example method shown in Fig. 2, the method 400 shown in Fig. 4 includes, at 414, obtaining a measurement signal comprising measurements of the sample 104 by the sensor unit 106. However, the method 400 proceeds, at 442, to process the measurement signal to compute values of a metric. Processing the measurement signal may comprise a combination of one or more techniques for computing the metric values as described below. In examples, the measurement signal may be processed by a metric module without relying on a machine learning model, such as a classifier or a regressor. For example, Fig. 3 shows the measurement signal 324 being processed by the metric module 340, which transfers the computed metric values for one or more portions of the measurement signal 324 to the classification module 336. Two example techniques of processing the measurement signal 324 to compute the metric values will now be described. In one example, processing the measurement signal at 442 may include computing an autocorrelation of the measurement signal 324. In such examples, the metric module 340 may compute an autocorrelation for one or more portions of the measurement signal 324. In particular, the autocorrelation of the measurement signal 324 may be computed as a function of time (assuming the measurements to be time-ordered or sequential). Based on the computed values of the autocorrelation of the measurement signal 324, portions of the measurement signal 324 may be associated with respective states of the nanopore 308. For example, it may be determined that computed values of the autocorrelation lie within a range typically exhibited when a polymer 316 is translocated through the nanopore 308. In such examples, portions of the measurement signal 324 exhibiting values in such a range may be determined to be associated with a state representative of the polymer 316 being translocated. In other examples, portions of the measurement signal 324 may be associated with other states of the nanopore 308. For example, it may be determined that the computed values of the autocorrelation lie outside a range associated with a nanopore state representing the polymer 316 being translocated through the nanopore 308. Autocorrelation of a process may be mathematically defined as the correlation between values of the process at two different times, e.g. at a lag. Regions of a signal representing the process in which data points exhibit similar values at the two different times, may have higher autocorrelation values. Likewise, regions in which the data points exhibit dissimilar values have lower autocorrelation values. In this manner, values of an autocorrelation function may indicate regions of similarity of the process at the two different times. Portions of the measurement signal 324 associated with the polymer 316 being translocated through the nanopore 308 may be expected to exhibit a reasonable degree of similarity due to the repeating sequence of polymer units included in the polymer 316. Thus, measurements associated with a nanopore state representing the polymer 316 being translocated may also be associated with a high autocorrelation value. In this manner, the autocorrelation may serve as a useful metric for deducing whether or not a given portion of the measurement signal 324 is associated with the polymer 316 being translocated through the nanopore 308 or not. Additionally, the inventors have found that the autocorrelation function is particularly effective in contexts where the complexity of the polymer being translocated through the nanopore 308 is low. For example, certain regions of a DNA polynucleotide may be rich in two nucleobases. In one example of clinical relevance, the region may be rich in Adenine (A) and Guanine (G) nucleobases. Measurements of such AG-rich regions may be difficult to distinguish from measurements of BLOCKED states of the nanopore, particularly with the use of hard-coded metrics. For example, BLOCKED states may exhibit a current variation within a reduced range of values akin to the variation exhibited by LOW-COMPLEXITY states of the nanopore. Thus, accurately determining whether a portion of the measurement signal 324 is associated with a LOW-COMPLEXITY state may mitigate dropouts in the coverage of the polymer 316 being sequenced. In low-complexity regions of polymers (such as the AG-rich regions of polynucleotide DNA), any two sections of the region show a higher degree of similarity, compared with arbitrary pairs of sections from a higher-complexity region involve more polymer units. In other words, the chances of an AG-rich region to be similar to itself at two different times is simply higher because there are fewer repeating units involved. Since the autocorrelation function exhibits accentuated values for regions of the measurement signal that are similar to their time-warped versions, low-complexity regions of the polymer 316 (such as AG-rich regions) exhibit higher values than higher-complexity regions. Thus, using the autocorrelation function to compute values of the metric to support classification has the additional benefit of enabling relatively accurate identification of low-complexity regions. The use of such techniques may similarly enable detection of HSAT3 regions in polynucleotides, and prevent flicking of the nanopore by confusing the measurements with BLOCKED states of the nanopore 308. In another example, processing the measurement signal as part of method 400 may include computing a number of crossings of a given value of the measurement signal. In such examples, the metric module 340 may process the measurement signal 324 to compute the number of a crossings as a function of time for a portion of the measurement signal 324. For example, the metric module 340 may compute the number of mean value crossings of the measurement signal 324 as a function of time. The number of crossings for a portion of the measurement signal 324 may be indicative of the state of the nanopore 308 associated with portion. In further examples, the mean value of the measurement signal 324 may be subtracted from the measured values (demeaning) prior to computing the number of crossings of the measurement signal 324 as a function of time. In this manner, a zero-crossing rate (ZCR) may be computed for different portions of the measurement signal 324. In examples, the portions associated with a ZCR in a range associated with the polymer 316 being translocated through the nanopore 308 may be classified as such, whereas the remaining portions may be associated with other states of the nanopore 308. In other examples, relatively more complex criteria may be used to associate a portion of the measurement signal 324 with a respective state of the nanopore 308 based on the computed values of the ZCR. Regardless of the method of association the portions with respective nanopore states, calculating a number of crossings (e.g. ZCR) of the measurement signal 324 may serve as an indicator of the nanopore state that may be quick to compute. Additionally, the number of crossings in a portion of the measurement signal 324 may be indicative of the rate of translocation of polymer units through the nanopore 308 for the portion of the measurement signal 324. For example, a ZCR of 100 counts / second may indicate that a region of the polymer 316 corresponding to the portion of the measurement signal 324 was translocated at the rate of approximately 100 polymer units per second. Where the polymer 316 is a polynucleotide, the ZCR may indicate the rate of translocation of nucleotides per unit time. By computing the duration of a given portion of the portion of the measurement signal 324 in time, the number of nucleotides may therefore be estimated. For example, if the ZCR for a portion lasting 10 seconds lies in range associated with a STRAND state and is on average found to be 200 counts per second, it may be estimated that the corresponding polynucleotide strand contained 2000 nucleotide units. In examples, it may be known that polynucleotide fragments in a sample 304 contain at least 5000 nucleotide units. In such examples, portions of the measurement signal 324 being associated with an estimated nucleotide count significantly less than 5000 bases may be associated with a NON-STRAND state. Similarly, sample preparation may be known to involve introduction of adapter polynucleotide molecules that may include a fixed number of bases, e.g. 2000 bases. Portions of the measurement signal 324 for which a ZCR based estimate of the nucleotide units is close to 2000 bases may be associated with an “Adapter” state of the nanopore 308, representative of one of the adapter molecules being translocated through the nanopore 308. In this manner, calculation of the ZCR value may provide further information, which may be relied upon in support of downstream processing. Whether by calculating the ACF or the ZCR of the measurement signal 324, values of a metric may be computed for the measurement signal 324. In some examples, these methods may be combined to compute multiple metrics that together serve as indicators of the nanopore state associated with each portion of the measurement signal 324. Regardless of the combination of ACF and ZCR adopted, metric values may be computed inexpensively for portions of the measurement signal 324 compared with the use of a machine learning model 330. In particular, the use of a machine learning model 330 may necessarily rely on computationally expensive matrix multiplications, which may be avoided by relying on the vectorial calculations involved in calculating either the ACF or the ZCR of the measurement signal 324. Similarly, computing metric values based on ACF and ZCF may also avoid resource-intensive training of the machine learning model 330. Such techniques may thus be useful in resource-poor contexts. When the availability of computational resources is not a concern, however, these techniques may be used in conjunction with a machine learning model 330. For example, ZCF and / or ACR may be used in a pre-filtering stage prior to processing the measurement signal 324 using a machine learning model 330. Alternatively, metrics computed using ACF and ZCR techniques may be provided to a machine learning model 330 as additional inputs alongside the measurement signal 324. Such inputs may enhance the information available to the machine learning model 330 for classification or regression. Thus, by combining ACF and ZCR techniques with the use of a machine learning model 330, the overall accuracy of read splitting may be enhanced. Fig. 4 shows that the method 400 proceeds, at 428, to determine that a portion of the measurement signal 324 is associated with a state of the nanopore 108. The associated state of the nanopore 308 may be one of a plurality of states as described previously in the context of Fig. 2. In examples, determining that the portion is associated with the state of the nanopore 308 may include identifying values from the computed values of the metric that exceed a threshold. For example, as described above, a threshold may be a minimum value known to be typically exceeded by measurements of polymer units of a polymer 316. Subsequently, the portion may be determined in dependence on the identified values. For example, a portion of the measurement signal 324 exhibiting values exceeding such a minimum may be associated with a state representative of the polymer 316 being translocated through the nanopore 308. Likewise, portions of the measurement signal 324 exhibiting values lower than the minimum value may be associated with other nanopore states. In examples, the threshold value may be part of a range known to be exhibited by polymer units being translocated through the nanopore 308. Upon associating the portion of the measurement signal 324 with the nanopore state, the method 400 proceeds by determining, at 434, whether to further analyse the portion of the measurement signal in dependence on the determination that the portion of the measurement signal is associated with the first state of the nanopore. Step 434 of method 400 is analogous to the step 234 of method 200. As described in that context, the method 400 may involve determining to further analyse the portion in dependence on determining that the portion is associated with a state representative of the polymer 316 being translocated through the nanopore 308. Portions associated with other nanopore states may be omitted from further analysis. Thus, metric values computed based on ACF and ZCR techniques may enable read splitting to be achieved for the measurement signal 324 in a computationally inexpensive manner. As described above in the context of method 200, further analysis of the measurement signal 324 in method 434 may involve sequencing one or more portions of the measurement signal 324. For example, where sequencing of polynucleotides is intended, portions determined to be associated with STRAND states of the nanopore 308 based on the computed metric values may be basecalled. In some examples, the computed values of the metric may be provided to a basecaller as additional information alongside the measurement values included in measurement signal 324. A basecaller may be trained to rely on such additional information as a means to further enhance the accuracy of basecalling. In general, a variety of actions may be taken in dependence on the nanopore state determined to be associated with a portion of the measurement signal 324. Fig. 5 shows a flow diagram representing operation decisions taken in dependence on processing a measurement signal obtained from a sensor unit. In particular, the decisions may vary for each portion of the measurement signal 324 in dependence on the nanopore states associated with the portion. For example, as described above in the context of methods 200 and 400, further analysis may be carried out for portions of the measurement signal 324 determined to be associated with nanopore states representative of the polymer 316 being translocated through the nanopore 308. As an example in the context of polynucleotide sequencing, Fig. 5 shows that the method 500 involves, at 544, deciding to further analyse a portion of a measurement signal 324 that is associated with a STRAND state of the nanopore 308. In such examples, further analysis may involve basecalling the portion as described previously. Fig. 5 also shows that the method 500 involves deciding to ignore, at 546, portions that are associated with an OPEN state of the nanopore 308, which is representative of the nanopore 308 not being occupied by the polymer 316. Ignoring such portions of the measurement signal 324 may result in greater computational economy of the method 500. Fig. 5 also illustrates decisions taken cases where processing of the measurement signal 324 results in associating the signal portion with a state representative of the polymer 316 being blocked in the nanopore 308. In such examples, the method 500 may involve instructing the sensor unit 306 to unblock the polymer. For example, Fig. 3 the classification module 332 providing an instruction to the sensor unit 306 based on associating the portion with the state 322, which is representative of a polynucleotide being blocked in the nanopore 308. In such contexts, the method may specifically involve providing instructions to the electrical circuitry in the sensor unit 106 in dependence on the read splitting performed at the classification module 332. For example, the classification module 332 may be capable of determining that a portion of the measurement signal is associated with a polymer being stuck in the nanopore 108, i.e. with the nanopore 108 being blocked and therefore inoperative. In consequence, the classification module 332 may instruct the sensor unit 106 to “flick” the nanopore 108 in order to eject the polymer from the nanopore 108 as described previously. Accurate read splitting conducted using the techniques described above may therefore lead to precise flicking protocols. For example, in some cases the read splitting may determine that the portion is associated with the polymer 316 being blocked temporarily within the nanopore 308. In such examples, the potential difference across the nanopore 308 may be reversed for a short period of time to temporarily dislodge the polymer 316, followed by reverting the potential difference to the original value to encourage translocation of the polymer 316 to be continued. In other examples, read splitting may determine that the portion is associated with the polymer 316 being permanently blocked within the nanopore 308. In such examples, it may be desirable to permanently eject the polymer 316 from the nanopore 308. By doing so, the basecalling workload of the system 100 may be dedicated to the regions of the measurement signal corresponding to STRAND states. Alternatively, the processing of BLOCKED state measurements may result in instructing the sensor unit to at least cause a change of state of the nanopore 308. In this manner, the nanopore 308 may be operated in states other than BLOCKED states, contributing to improved sequencing yields. In either case, measurements corresponding to the BLOCKED state may be processed to determine appropriate responsive actions. By taking appropriate actions to resolve the block, the nanopore 308 may be operated to spend less time in a terminally blocked state, thereby improving the chances of the nanopore 308 being in a STRAND state, and consequently enhancing the overall throughput of the system 100. It is noted while only three states of the nanopore 308 are illustrated in Fig. 3, these are not by any means an exhaustive list of nanopore states. Through precise control of flicking behaviour, the time spent by a nanopore 308 in a state associated with translocating a polymer 316 may be enhanced, and likewise the time spent in a blocked state may be reduced. By achieving improved read splitting, computational efforts may be concentrated on processing the portions of the measurement signal that are associated with polymers, thereby improving the sequencing yield of the flow cell 102. Similarly, with improved flicking behaviour, the operation of the flow cell 102 may be dedicated to translocating and measuring polymers, also leading to improved yields. Methods described herein relate to operating a polymer analysis system based on improved techniques for read splitting and / or flicking behaviour. Overall, the methods described give rise to improved sequencing throughput, but also have other technical advantages that will be described where relevant. The decisions shown apply to both methods 200 and 400 that have been described thus far. Fig. 6A and Fig. 6B illustrate two scenarios in which the methods described thus far prove to be effective. Fig. 6A shows a first example of a measurement signal obtained from a sensor unit and a corresponding graph showing computed metric values. In particular, the top panel of Fig. 6A shows an illustrative measurement signal 548 obtained from the sensor unit during translocation of a polynucleotide through the nanopore 308. Measurements included in the measurement signal show systematic fluctuations corresponding to interactions between respective groups of polymer units and the nanopore 308. However, the measurements also show an overall drift in the downward direction. Although the measurements appear to fluctuate within a similar range of values at all times, measurements at later times (right hand side) appear to exhibit lower values than the measurements at earlier times (left hand side). The measurement signal 548 therefore serves to illustrate the current drift phenomenon discussed earlier. The top panel of Fig. 6A also indicates a range of measurement values (using dashed horizontal lines in the top panel) typically expected for a polynucleotide translocating through the nanopore 308. In particular, the typical range of values is shown to be bounded by a lower threshold value 550 and an upper threshold value 552. The bottom panel of Fig. 6A shows metric values computed for the measurement signal. In particular, dashed lines indicate metric values 554 computed using the hard-coded criterion of measurement values lying within the fixed typical range associated with a polynucleotide being translocated through the nanopore 308. The solid line indicates metric values 556 computed using the methods described hereinbefore. It is noted that the computed metric values 554 and 556 are illustrative only and may be computed either using a machine learning model 330 or using a combination of the ACF or ZCR methods described previously. Fig. 6A shows that measurements obtained at earlier times (left hand side of the top panel) clearly exhibit values within the typical range of values bounded by the threshold values 550 and 552. As a result, the corresponding values 554 of the hard-coded metric exhibit values are found to be greater than a critical value 558 (solid horizontal line in the bottom panel). However, at later times the measurements exhibit lower values due to the overlaid drift as a function of time (right hand side of top panel). This results in the corresponding hard-coded metric values 554 shifting below the critical value 558 at later times (right hand side of bottom panel). Due to the change in metric values from above to below the critical value, methods relying on the hard-coded metric values 554 for read splitting may estimate later portions of the measurement signal 548 to be associated with the polynucleotide not being translocated through the nanopore 308. This is erroneous as the entirety of the measurement signal 548 in fact corresponds to the translocation of a polynucleotide. The metric values 556 computed using the methods described herein before do not exhibit the same drawback. In particular, the solid line 556 is shown to remain above the critical value 558 through the experiment despite the drift in measurements. This is attributable to the fact that the techniques discussed hereinbefore can account for complex dependencies between the measurement data and the nanopore states. In particular, the machine learning model 330 may account for non-linear dependencies between the measurements, measurement time (as well as any other computed metrics) and the state of the nanopore 308. Similarly, the zero crossing rate is indicative of the number of crossings of the measurement signal 548 of a mean value, rather than relying on a precise (hard-coded) range of values exhibited by the measurements at all times. Since portions of the measurement signal 548 correspond to a STRAND state, they also exhibit a similar rate of crossing at earlier and later (drifted) values of the measurements. In consequence, the corresponding metric values 556 may record all measurements as being above the critical value 568. Likewise, the ACF technique discussed previously may also result in a similar prediction. This is because the metric values 556 computed using the ACF may be indicative of a similarity between measurements in the measurement signal 548 at two different times. Since measurements at earlier times in the measurement signal 548 exhibit similar fluctuations as measurements at later times, the corresponding metric values 556 calculated using ACF may remain above a critical value 558. In this manner, metric values 556 computed using methods described hereinbefore may enable accurate read splitting, allowing downstream processing tasks to be performed efficiently. Fig. 6B shows a second example of a measurement signal obtained from a sensor unit and a corresponding graph showing computed metric values. In particular, the top panel of Fig. 6B shows a measurement signal 554 containing measurements of a polynucleotide with high- and low-complexity regions obtained during translocation. The portion 562 of the measurement signal 560 is associated with a high-complexity region of the polynucleotide, and thereby exhibits a wide range of fluctuations. The portion 564 of the measurement signal 560 is associated with a low-complexity region of the polynucleotide including mainly A and G nucleobases, and thereby exhibits a narrower range of fluctuations. In particular, the fluctuations exhibited by measurements of the low-complexity region are below a threshold value 566. Fog. 6B shows (in the bottom panel) computed metric values for the measurement signal 560. The dashed lines 568 correspond to metric values computed using a hard-coded constraint requiring that the fluctuations exhibited by the measurements attain values on either side of the threshold value 566. As a result, the hard-coded metric values 568 are seen to be greater than a critical value 570 for the earlier portion 562 of the measurement signal 560, whereas the values 568 are seen to be lower than the critical value 570 for the portion 564 corresponding to the low-complexity region of the polynucleotide. In consequence, read splitting based on the hard-coded metric values may result in associating the portion 564 with a different state of the nanopore 308, rather than correctly associating it with a STRAND state. In some examples, the portion 564 may be associated with a BLOCKED state of the nanopore 308. The solid line 572 in the bottom panel of Fig. 6B shows metric values computed using the techniques described hereinbefore. The computed metric values 572 are seen to exhibit values greater than the threshold value 570 for the entire measurement signal 560. This may be because the machine learning model 330 is capable of accurately determining the state associations for different types of polymer classes, including high-and low-complexity classes. Similarly, as described previously, the ZCR and ACF techniques are also capable of accurately associating low-complexity regions to STRAND states rather than other states of the nanopore 308. This is because, for instance, a low-complexity AG-rich region of the polynucleotide 316 may exhibit a similar zero crossing rate obtained after subtracting a mean value from the measurement signal 560, even though the values of low-complexity portion 564 of the measurement signal may be lower than a threshold value 570. Likewise, the ACF may attribute values greater than the critical value for each of the portions 562 and 564 of the measurement signal 560. This is because the measurements within either of these portions may be highly similar at two different times, and therefore the associated ACF for each of these portions may be above the critical value 570. In this manner, the methods described hereinbefore may enable accurate downstream processing for polymers contain high-and low-complexity regions. The above embodiments are to be understood as illustrative examples of the invention. Further embodiments of the invention are envisaged. For example, embodiments may include a reporting system such as a display for indicating the associations of different portions of the measurement signal with respective nanopore states. Such a display may be updated at runtime in order to allow users operating the polymer analysis system to take workflow management decisions. In some examples, the user may be able to instruct the polymer analysis system to stop operating a given sensor unit or a nanopore whose measurements are consistently associated with a BLOCKED state. In such examples, the user may be able to mark the nanopore for inspection and repair. More generally, the display may indicate actions that a user may take to optimise the operation of the polymer analysis system. It is to be understood that any feature described in relation to any one embodiment may be used alone, or in combination with other features described, and may also be used in combination with one or more features of any other of the embodiments, or any combination of any other of the embodiments. Furthermore, equivalents and modifications not described above may also be employed without departing from the scope of the invention, which is defined in the accompanying claims.
Claims
1. A computer-implemented method of processing a measurement signal comprising measurements of a sample by a sensor unit comprising a nanopore, the method comprising:analysing the measurement signal using a machine learning model to determine that a portion of the measurement signal is associated with a first state of the nanopore, the first state being one of a plurality of states of the nanopore; anddetermining whether to further analyse the portion of the measurement signal in dependence on the determination that the portion of the measurement signal is associated with the first state of the nanopore.
2. The method of claim 1, wherein the machine learning model is a classifier, and wherein analysing the measurement signal comprises classifying the measurement signal using the machine learning model to determine that the portion is associated with the first state.
3. The method of either claim 1 or claim 2, wherein:the measurement signal is associated with timestamps indicating times that respective measurements were taken by the sensor unit; andanalysing the measurement signal comprises processing the timestamps in combination with the measurement signal using the machine learning model.
4. The method of any preceding claim, wherein analysing the measurement signal comprises processing the measurement signal by the machine learning model to determine values of a metric for measurements of the sample within the measurement signal,wherein determining that the portion of the measurement signal is associated with the first state of the nanopore is based at least in part on the determined values of the metric.
5. A computer-implemented method of processing a measurement signal comprising measurements of a sample by a sensor unit comprising a nanopore, the method comprising:processing the measurement signal to determine values of a metric, the processing comprising at least one of:computing an autocorrelation of the measurement signal; or computing a number of crossings of a given value of the measurement signal;determining, using the determined values of the metric, that a portion of the measurement signal is associated with a first state of the nanopore, the first state being one of a plurality of states of the nanopore; anddetermining whether to further analyse the portion of the measurement signal in dependence on the determination that the portion of the measurement signal is associated with the first state of the nanopore.
6. The method of claim 5, comprising calculating a mean value for at least part of the measurement signal,wherein the metric comprises a number of crossings of the mean value for said at least part of the measurement signal.
7. The method of either claim 5 or claim 6, wherein the first state is representativeof a polymer being translocated through the nanopore, the method comprising: estimating a count of polymer units in the portion in dependence on the computed values of the metric; andassociating the portion of the measurement signal with the first state in dependence on the estimated count.
8. The method of any of claims 4 to 7, wherein determining that the portion is associated with the first state of the nanopore comprises:identifying values from the determined values of the metric that exceed a threshold; anddetermining the portion in dependence on the identified values.
9. The method of any of claims 4 to 8, comprising providing the measurement signal and the determined values of the metric to a basecaller.
10. The method of claim 9, wherein the first state is representative of a polymer being translocated through the nanopore, and wherein the metric represents an estimated dwell time of the polymer within the nanopore.
11. The method of claim 9, wherein the metric represents confidence in a given set of measurements being associated with the polymer.
12. The method of claim 11, wherein the metric comprises a confidence value indicating confidence in the given set of measurements being associated with a first polymer class, the first polymer class being one of a plurality of polymer classes.
13. The method of claim 12, wherein the plurality of polymer classes include at least one ofa high-complexity polymer,a low-complexity polymer,a polypeptide, anda polynucleotide adapter, andwherein the method comprises further analysing the portion of the measurement signal in dependence on the confidence value.
14. The method of any preceding claim, comprising determining a length of the portion of the measurement signal,wherein determining whether to further analyse the portion of the measurement signal is further dependent on the determined length.
15. The method of any preceding claim, wherein the first state is representative of a polymer being translocated through the nanopore, the method comprising determining to further analyse the portion of the measurement signal.
16. The method of claim 14, wherein further analysing the portion comprises basecalling the portion to estimate a sequence of polymer units of the polymer.
17. The method of any of claims 1 to 6, wherein the first state is representative of a polymer being blocked in the nanopore, the method comprising instructing the sensor unit to unblock the polymer in the nanopore.
18. The method of claim 16, wherein unblocking the polymer comprises ejecting the polymer from the nanopore by reversing a voltage applied across the nanopore.
19. The method of any of claims 1 to 6, wherein the first state is representative of the nanopore not being occupied by a polymer, the method comprising determining not to further analyse the portion of the measurement signal.
20. A data processing system comprising means for carrying out the method of any of claims 1 to 19.
21. A computer program product comprising instructions which, when the program product executed by a computer, cause the computer to carry out the method of any of claims 1 to 19.
22. A method comprising:providing a sample; andperforming the method of any of claims 1 to 19 to process a measurement signal comprising measurements of the sample by a sensor unit comprising a nanopore.
23. An apparatus comprising:a sensor unit comprising a nanopore; anda data processing system arranged to process a measurement signal comprising measurements of a sample by the sensor unit in accordance with the method of any of claims 1 to 19.
24. A kit for use in processing a measurement signal comprising measurements of a sample, the kit comprising the apparatus of claim 23.
Citation Information
Patent Citations
ViewUS2018/0037948A1onEspacenetopensinnewtab
ViewUS10102338B2onEspacenetopensinnewtab
ViewWO2020/168286A1onEspacenetopensinnewtab
