FAST-NA for threat detection in high-throughput sequencing
The system uses sequence filters and probabilistic data structures to detect and classify malicious genetic sequences in real-time, addressing the challenge of distinguishing between benign and malicious organisms in nucleic acid sequencing.
Patent Information
- Application Number
- JP2022562675
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-04-22
- Filing Date
- 2021-02-22
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2041-02-22
AI Technical Summary
Current nucleic acid sequencing technologies struggle to detect and classify malicious genetic sequences, such as those of pathogens, due to advances in oligo-based assembly and editing techniques that allow for the assembly of small DNA segments, making it difficult to distinguish between benign and malicious organisms.
The system employs sequence filters to identify and classify genetic sequences by breaking down known malicious and benign sequences into short fragments, using probabilistic data structures like Bloom filters to compare and store signatures, allowing real-time or near-real-time detection of malicious organisms.
Enables rapid identification and classification of malicious genetic sequences, reducing the risk of synthesizing harmful organisms by flagging suspicious sequences for further investigation, and providing metadata for analysis, thus enhancing security in nucleic acid synthesis.
Smart Images

Figure 0007590075000001 
Figure 0007590075000002 
Figure 0007590075000003
Abstract
Description
[Technical field]
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority under 35 U.S.C. §119(e) to U.S. Provisional Application No. 63 / 013,872, filed April 22, 2020, entitled "FAST-NA FOR DETECTION AND DIAGNOSTIC TARGETING," and U.S. Provisional Application No. 63 / 013,875, filed April 22, 2020, entitled "FAST-NA FOR THREAT DETECTION IN HIGH-THROUGHPUT SEQUENCE," each of which is incorporated by reference in its entirety herein.
[0002] Technical Field This application relates generally to detecting specific genetic sequences, and more particularly, in one aspect, to systems and methods for using sequence filters to identify and / or classify malicious genetic sequence fragments. [Background technology]
[0003] Laboratories are now able to use nucleic acid sequence information to manufacture deoxyribonucleic acid (DNA) and other sequences. In an exemplary scenario, a customer provides the nucleotides in a genetic sequence (in a format as simple as an electronic text file) to a laboratory, which synthesizes (i.e., manufactures) the sequence for delivery to the customer. This technology raises the fear of malicious actors covertly requesting the synthesis of malicious organisms. Diseases such as influenza and anthrax could be effectively "mail-ordered," thereby posing a public health risk. To prevent such a scenario, laboratories offering such synthesis services typically test genetic sequences provided by customers to ensure that the sequences are not associated with malicious organisms.
[0004] Current technology can recognize sequences as short as about 200 base pairs. However, recent advances in oligo-based assembly and editing, such as the Clustered Regularly Interspaced Short Palindromic Repeat (CRISPR) machinery, make it possible to "clip and splice" together small segments of DNA. Thus, unscrupulous customers could avoid detection by embedding parts of a bad organism into the DNA sequences of multiple good organisms, or otherwise synthesizing the bad organism in small pieces. Pathogenic sequences from these short or hybrid DNA sequences could be reassembled into the bad organism after they are synthesized and delivered.
[0005] A sample of genetic material may include a wide variety of genetic material, including pathogenic genetic material. The sample may be analyzed to identify one or more pathogens present in the sample. For example, the sample may be analyzed to determine whether a pathogen of interest is present in the sample. The sample may be sequenced by a "sequencer" configured to analyze the sample and output information indicative of the genetic sequence of the sample. The genetic sequence may be analyzed to determine whether a pathogen of interest is present in the sample. Summary of the Invention [Means for solving the problem]
[0006] Aspects and embodiments are directed to devices and methods for identifying target organism "signatures" (relatively short fragments of genetic sequence present in a target organism but not present in similar but non-target organisms). What is considered a "target organism" and what distinguishes a target organism from a "non-target organism" may be controlled or selected by a user. For example, a target organism may be a malicious organism such as the SARS-CoV-2 coronavirus, and a non-target organism may be a benign organism such as other non-malicious coronaviruses. In other examples, a target organism signature may refer more broadly to the signature of any target organism that may be desirable to distinguish from a non-target organism. That is, a target organism may be a malicious organism, but it should be understood that the phrase "target organism" is broader than a malicious organism. For example, a target organism may be an organism that belongs to a particular species, genus, or other biological classification, and a non-target organism may be an organism that does not belong to that particular biological classification. Other examples of target criteria are within the scope of the present disclosure. In various examples, the target organism or target genetic sequence may alternatively or additionally be referred to as a "genetic sequence of interest" or a "pathogen of interest," regardless of how the sequence or pathogen is otherwise classified, such as as "benign," "malignant," etc.
[0007] However, for the sake of illustration, examples are given in which the target organism is a malicious organism and the target signature is a malicious genetic sequence fragment. These examples are provided for illustrative purposes and are not intended to be limiting. Detection of such a signature in a sequence to be analyzed can indicate, with a level of certainty, that the sequence contains malicious genetic code. For example, if the sequence is a sequence requested for synthesis, synthesis of the sequence can be rejected or postponed until further investigation and review is completed. It should be understood that, although in some examples, the principles of the present disclosure are applicable to synthesis procedures (e.g., to aid in the decision as to whether or not to perform a synthesis procedure), the principles of the present disclosure are not limited to examples involving synthesis. Thus, no limitation is implied by the examples involving synthesis, which are provided merely for illustrative purposes.
[0008] Such signatures can also be used to classify sequences according to the type of organism (malicious or not) for which the sequence contains genetic information. In another example, a sample may be analyzed to determine whether a sequence contains a target genetic code to identify a target organism in the sample. For example, a sample may be analyzed to determine whether the sample contains the SARS-CoV-2 coronavirus.
[0009] To identify a signature of a malicious organism, the sequence of a known malicious organism and the sequences of one or more known benign organisms may be used. Each sequence is broken down into relatively short fragments, and the fragments of the malicious organism are compared to the fragments of the benign organism. For a more efficient comparison, the fragments of the benign organism may be placed into a probabilistic data structure, such as a Bloom filter. If a match is found, i.e., if the fragment of the malicious organism is also present in the benign organism, then the fragment of the malicious organism is not a suitable signature for the malicious organism. On the other hand, if the signature cross section of the malicious organism is known to be present only in the malicious organism, then the fragment of the malicious organism may be a suitable signature. The suitable signature may be stored in a database of malicious signatures along with metadata about the fragment or the corresponding malicious organism, including the species of the organism, an identifier of the sample from which the fragment was taken, and / or the location of the fragment within the sample.
[0010] In some examples, the set of signatures may be filtered to identify a smaller set of one or more universal signatures. For example, a group of variants of a known malicious organism, such as the sequences of several variants of the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) virus that causes coronavirus disease 2019 (COVID-19), may be analyzed to obtain one or more target signatures. As discussed above, each of the target signatures may be identified as a signature because it is not present in any of the sequences of the one or more known benign organisms and thus may uniquely identify the malicious organism. The group of target signatures may then be filtered by identifying those target signatures that are present in all variants of the group of variants. The target signatures in this subset may be referred to as "universal signatures" since each signature is universally present in all variants of that malicious organism. For example, consider a group of variants of the SARS-CoV-2 virus. A universal signature is a signature that is present in the sequence of each variant of the SARS-CoV-2 virus and is absent in the sequence or sequences of the benign organisms used to generate the universal signature. The universal signature may optionally be further processed and stored in a signature database for subsequent analysis.
[0011] In one approach, an unknown sequence (e.g., provided by a customer) can be tested by comparison to a signature database that can quickly determine whether the sequence contains one or more signature fragments, thereby identifying the sequence as potentially malicious. Metadata stored with the fragments may be used to aid or refine the identification and determine a confidence level in the identification, or may be provided to other systems or users for further analysis. In a second approach, the sequence being tested can be compared to multiple such filters, each of which contains signatures for a particular category of organisms. For example, one filter may identify influenza signature fragments and another filter may identify anthrax signature fragments. For more efficient classification, each filter of signature fragments may be placed into a probabilistic data structure, such as a Bloom filter. In this way, the sequence can be classified according to one or more types of organisms for which it contains genetic information.
[0012] In some examples, unknown sequences output by a sequencer (e.g., DNA sequences output by a DNA sequencer) based on one or more organism samples may be analyzed before the sequencer finishes sequencing the samples. The analysis may be performed in real-time or near real-time with the operation of the sequencer. That is, the analysis may be performed while the sequencer is still sequencing the samples. In some examples, the analysis may be performed at a rate approximately equal to or greater than the rate at which the sequencer sequences the samples, as measured in megabytes (MB) per minute. Thus, as used herein, "real-time" or "near real-time" analysis refers to analysis performed on a test sequence as it is output by a sequencer, without any intentional delay between the test sequence being output by the sequencer and the test sequence being analyzed that is not incidental to the analysis procedure. In some examples, an analysis may be considered "real-time" or "near real-time" as long as the analysis is performed at a rate approximately equal to or greater than the rate at which the sequencer sequences the samples, as measured in MB / min. In some examples, analysis may be performed and / or completed on a particular portion of a gene sequence within milliseconds of the gene sequence being initially sequenced by a sequencer, before the sequencer has finished sequencing the sample. In other examples, analysis may be performed and / or completed on a particular portion of a gene sequence within different times, such as within nanoseconds, microseconds, seconds, etc.
[0013] A record may be kept of the number of times the target is detected in the sequence. In some instances, the entire sequence output by the sequencer may be analyzed, so there may be many opportunities to identify a malicious fragment that indicates the target is represented by that sequence. The probability that each of one or more threats is present in the sample may be determined based on the number of times each threat is identified. For example, a threat that is only identified once in the sequence may be assigned a low probability, while a threat that is identified more than 100 times in the sequence may be assigned a high probability.
[0014] The systems and methods described herein are not limited to identifying and / or classifying malicious organisms. For example, in some applications, genetic sequences (e.g., from non-malicious organisms) may be compared against a signature database to identify species, taxons, or other categories of organisms.
[0015] According to one aspect, a method for identifying regions of a malicious organism sequence is provided, the method comprising: identifying a plurality of benign fragments derived from a first sequence obtained from at least one benign organism; extracting a plurality of candidate signature fragments from a second sequence obtained from the malicious organism; for each of the plurality of candidate signature fragments, determining whether the candidate signature fragment matches at least one of the plurality of benign fragments; and in response to the candidate signature fragment not matching the at least one of the plurality of benign fragments, identifying the candidate signature fragment as a malicious signature fragment.
[0016] In an embodiment, the method includes determining whether a malicious signature fragment is present in at least one test sequence. In a further embodiment, the method includes, for a plurality of malicious signature fragments present in the at least one test sequence, determining a common feature of the plurality of malicious signature fragments. In yet another embodiment, determining a common feature of the plurality of malicious signature fragments is performed with reference to metadata for at least one fragment of the plurality of malicious signature fragments. In a further embodiment, the metadata includes at least one of an identifier of the genus of the organism from which the fragment was obtained, an identifier of the species of the organism from which the fragment was obtained, and a location on the second sequence from which the fragment was generated.
[0017] In another embodiment, the plurality of benign fragments and the candidate signature fragments are one of DNA fragments, RNA fragments, and amino acid fragments. In another embodiment, the plurality of benign fragments are arranged in a probabilistic data structure. In a further embodiment, the probabilistic data structure is one of a Bloom filter and a search tree.
[0018] In one embodiment, identifying the plurality of benign fragments comprises extracting the plurality of benign fragments from the first sequence obtained from at least one benign organism. In another embodiment, the at least one benign organism is a non-malicious strain of an organism having at least one malicious strain. In yet another embodiment, the at least one benign organism belongs to a genus having at least one malicious organism.
[0019] In another embodiment, the method includes predicting a minimum number of benign fragments to be included in the plurality of benign fragments, the minimum number being sufficient to provide a false positive rate below a threshold, the false positive rate being the rate at which candidate signature fragments identified as malicious fragments are present in sequences of benign organisms. In a further embodiment, the minimum number of benign fragments is selected in relation to the type of malicious organism.
[0020] In one embodiment, the plurality of benign fragments are a plurality of n-long subsequences of a first sequence, and the malicious fragment is an n-long subsequence that is not among the plurality of n-long subsequences.
[0021] In another embodiment, the plurality of candidate signature fragments comprises a first plurality of n-long subsequences of the sequence, each of the first plurality of n-long subsequences beginning at a different position in the sequence, and the plurality of benign subsequences comprises a second plurality of n-long subsequences of a known benign sequence, each of the second plurality of n-long subsequences beginning at a different position in the known benign sequence. In another embodiment, the malicious fragment is a genetic sequence of a pathogen.
[0022] According to another aspect, a system is provided that includes a benign fragment database configured to store a plurality of benign fragments from a first sequence obtained from at least one benign organism, and a processor configured to extract a plurality of candidate signature fragments from a second sequence obtained from a malignant organism, for each of the plurality of candidate signature fragments, determine whether the candidate signature fragment matches at least one of the plurality of benign fragments, and in response to the candidate signature fragment not matching the at least one of the plurality of benign fragments, identify the candidate signature fragment as a malignant signature fragment.
[0023] According to another aspect, a method for classifying biological sequences is provided, the method comprising: generating a first plurality of sequence fragments from a first plurality of organisms having a first characteristic; generating a second plurality of sequence fragments from a second plurality of organisms having a second characteristic; identifying a plurality of benign sequence fragments; and filtering the first plurality of sequence fragments and the second plurality of sequence fragments to remove at least one of the plurality of benign sequence fragments.
[0024] According to one embodiment, the method includes determining whether a test sequence is present in the first plurality of sequence fragments, identifying the test sequence as having the first characteristic in response to the test sequence being present in the first plurality of sequence fragments, determining whether the test sequence is present in the second plurality of sequence fragments, and identifying the test sequence as having the second characteristic in response to the test sequence being present in the second plurality of sequence fragments.
[0025] In another aspect, the first plurality of sequence fragments, the second plurality of sequence fragments, and the plurality of benign sequence fragments are one of DNA fragments, RNA fragments, and amino acid fragments.
[0026] According to yet another embodiment, the first plurality of sequence fragments are arranged in a first probabilistic data structure and the second plurality of sequence fragments are arranged in a second probabilistic data structure. According to a further embodiment, the first probabilistic data structure and the second probabilistic data structure are one of a Bloom filter and a search tree, respectively.
[0027] According to yet another embodiment, the first trait identifies a first class of pathogen and the second trait identifies a second class of pathogen.
[0028] In accordance with at least one example, a method of analyzing sequencer output in real time is provided, the method including the steps of identifying a group of genetic targets; obtaining a plurality of target signature fragments in response to identifying the group of genetic targets, each target signature fragment being derived from a genetic sequence of a respective genetic target in the group of genetic targets; receiving a plurality of portions of a test sequence output by a sequencer that sequences a sample in real time; and determining a sequence of at least one of the plurality of target signature fragments in real time or near real time as the sequencer sequences the sample. The method includes determining whether a target signature fragment is present in at least one portion of the test sequence among the plurality of portions of the test sequence; for each genetic target of the group of genetic targets, determining a probability that each genetic target is present in the sample based at least on the determination whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of the test sequence among the plurality of portions of the test sequence; and outputting an analysis of the sample, the analysis indicating a respective probability that each genetic target is present in the sample.
[0029] In various examples, the plurality of portions of the test sequence includes a first portion and a second portion, and the first portion of the test sequence output by the sequencer is received before the second portion of the test sequence is generated by the sequencer. In at least one example, the method includes filtering the plurality of portions of the test sequence before determining whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of the test sequence of the plurality of portions of the test sequence. In various examples, filtering the plurality of portions of the test sequence includes removing low quality portions of the test sequence from the plurality of portions of the test sequence. In at least one example, the low quality portions of the test sequence include portions of the test sequence having more than a threshold number of consecutively repeated nucleobases.
[0030] In various examples, the method includes determining, for each genetic target of the group of genetic targets, a count value indicating the number of times at least one of the respective target signature fragments of the respective plurality of target signature fragments corresponding to the respective genetic target is determined to be present in the test sequence. In at least one example, determining, for each genetic target of the group of genetic targets, a respective probability that the respective genetic target is present in the sample is based on the respective count value of the respective genetic target. In various examples, the respective probability that the respective genetic target is present in the sample increases as the respective count value of the respective genetic target increases. In at least one example, the method includes determining whether the respective count value of the respective genetic target exceeds a respective threshold value, and determining that the respective genetic target is present in the sample based on the respective count value exceeding the threshold value.
[0031] According to at least one example, a system for analyzing sequencer output in real time is provided, the system having a memory, at least one database configured to store target signature fragments, and at least one processor coupled to the memory and the at least one database. The processor is configured to perform the steps of: identifying a group of genetic targets; obtaining a plurality of target signature fragments from the at least one database in response to identifying the group of genetic targets, each target signature fragment being derived from a genetic sequence of a respective genetic target of the group of genetic targets; receiving a plurality of portions of a test sequence output by a sequencer that sequences the sample in real time, and determining, in real time or near real time as the sequencer sequences the sample, whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of a test sequence of the plurality of portions of a test sequence; determining, for each genetic target of the group of genetic targets, a respective probability that each genetic target is present in the sample based at least on the determination whether at least one target signature fragment of the plurality of target signature fragments is present in the at least one portion of a test sequence of the plurality of portions of a test sequence; and outputting an analysis of the sample, the analysis indicating a respective probability that each genetic target is present in the sample.
[0032] In various examples, the plurality of portions of the test sequence includes a first portion and a second portion, and the first portion of the test sequence output by the sequencer is received before the second portion of the test sequence is generated by the sequencer. In at least one example, the system further includes the sequencer. In various examples, the at least one processor is further configured to filter the plurality of portions of the test sequence before determining whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of the test sequence of the plurality of portions of the test sequence. In at least one example, filtering the plurality of portions of the test sequence includes removing low quality portions of the test sequence from the plurality of portions of the test sequence. In various examples, the low quality portions of the test sequence include portions of the test sequence having more than a threshold number of consecutively repeated nucleobases.
[0033] In at least one example, the at least one processor is further configured to determine, for each genetic target of the group of genetic targets, a count value indicating the number of times at least one respective target signature fragment of the respective plurality of target signature fragments corresponding to the respective genetic target is determined to be present in the test sequence. In various examples, determining, for each genetic target of the group of genetic targets, a respective probability that the respective genetic target is present in the sample is based on the respective count value of the respective genetic target. In at least one example, the respective probability that the respective genetic target is present in the sample increases as the respective count value of the respective genetic target increases. In various examples, the at least one processor is further configured to determine whether the respective count value of the respective genetic target exceeds a respective threshold value, and determine that the respective genetic target is present in the sample based on the respective count value exceeding the threshold value.
[0034] According to various examples, a non-transitory computer readable medium is provided having stored thereon a sequence of computer executable instructions for analyzing output of a sequencer in real time, the sequence of computer executable instructions including: identifying a group of genetic targets; obtaining a plurality of target signature fragments in response to identifying the group of genetic targets, each target signature fragment being derived from a genetic sequence of a respective genetic target of the group of genetic targets; and receiving a plurality of portions of a test sequence output by a sequencer that sequences the sample in real time, in or near real time as the sequencer sequences the sample. The method includes instructions to perform the steps of: determining, at a time, whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of a test sequence of the plurality of portions of a test sequence; determining, for each genetic target of a group of genetic targets, a respective probability that each genetic target is present in the sample based at least on the determination of whether at least one target signature fragment of the plurality of target signature fragments is present in the at least one portion of a test sequence of the plurality of portions of a test sequence; and outputting an analysis of the sample, the analysis indicating a respective probability that each genetic target is present in the sample.
[0035] Further aspects, embodiments, and advantages of these exemplary aspects and embodiments are discussed in detail below. The embodiments disclosed herein can be combined with other embodiments in any manner consistent with at least one of the principles disclosed herein. References to "an embodiment," "some embodiments," "alternative embodiments," "various embodiments," "one embodiment," and the like are not necessarily mutually exclusive and are intended to indicate that a particular feature, structure, or characteristic described may be included in at least one embodiment. Appearances of such terms in this specification do not necessarily all refer to the same embodiment. [Brief description of the drawings]
[0036] Various aspects of at least one embodiment are described below with reference to the accompanying drawings, which are not intended to be to scale. The drawings are included to provide illustration and further understanding of the various aspects and embodiments, and are incorporated in and constitute a part of this specification, but are not intended as a definition of the outer boundaries of the disclosure. In the drawings, each identical or nearly identical component shown in various drawings is represented by a similar numeral. For clarity, not every component is labeled in every figure. [Figure 1] FIG. 1 is a block diagram of a computer system for identifying signature fragments and comparing them to test sequences according to aspects of the present disclosure. [Figure 2A] 1 illustrates extraction of benign fragments according to aspects of the present disclosure. [Figure 2B] 1 illustrates the extraction of candidate signature fragments according to aspects of the present disclosure. [Diagram 3] FIG. 2 is a flow diagram of an example process for identifying a signature fragment according to aspects of the present disclosure. [Figure 4] FIG. 1 is a flow diagram of an example process for testing a test array according to aspects of the present disclosure. [Diagram 5] FIG. 1 is a block diagram of another computer system for identifying signature fragments and comparing them to test sequences according to aspects of the present disclosure. [Figure 6] FIG. 2 is a flow diagram of an example process for identifying a signature fragment according to aspects of the present disclosure. [Figure 7] FIG. 1 is a flow diagram of an example process for testing a test array according to aspects of the present disclosure. [Figure 8] FIG. 1 is a block diagram of an example computer system upon which aspects and embodiments of the present disclosure may be implemented. [Figure 9] FIG. 1 is a diagram of several sets of target signature fragments according to aspects of the present disclosure. [Figure 10]FIG. 13 is a flow diagram of an example process for filtering signature snippets according to aspects of the present disclosure. [Figure 11] FIG. 13 is an example flow diagram for analyzing the output of a sequencer, according to an example. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0037] Systems and methods for identifying and classifying genetic sequences are described. For example, genetic features of malicious organisms (e.g., pathogens such as anthrax or influenza) may be isolated. Such a malicious organism signature may be a fragment of a genetic sequence that is present in the malicious organism but not in related benign organisms, thereby uniquely identifying the malicious organism sequence. Once such a malicious organism signature is identified, a test sequence of unknown composition can be compared to the malicious organism signature to quickly determine whether the test sequence contains the malicious organism signature. If so, the test sequence may be flagged for further investigation and / or identified as containing malicious sequence information.
[0038] Various approaches may be used to identify and / or classify malicious sequences. According to one approach, a benign fragment database is populated with sequences derived from known benign organisms. The sequences may represent deoxyribonucleic acid (DNA) sequences, ribonucleic acid (RNA) sequences, other nucleic acid sequences, amino acid sequences, etc. The benign fragment database may be arranged as a probabilistic data structure, such as a Bloom filter, and benign organisms may be selected for their similar structure or classification with malicious organisms of interest.
[0039] The system is "trained" by identifying one or more signature fragments for a particular malicious organism. In the training process, a sequence from a malicious organism is parsed into candidate signature fragments. A benign fragment database is then examined to determine whether each candidate signature fragment is present. If the candidate signature fragment is present in the benign fragment database, the candidate signature fragment is not a proper signature fragment, i.e., it is present in malicious and benign organisms alike and is therefore not useful in identifying malicious organisms. On the other hand, if the candidate signature fragment is not present in the benign fragment database, the candidate fragment may be a malicious signature fragment for use in identifying malicious organisms. That is, the presence of that malicious signature fragment in a test sequence means that the test sequence is not derived from any of the benign organisms represented in the benign fragment database. The malicious signature fragments may then be compiled into the malicious signature database as part of the training process. Metadata about the fragment and / or the corresponding malicious organism may also be stored, including the species of the organism, an identifier for the sample from which the fragment was taken, and the location of the fragment within the sample.
[0040] After the training process is complete, the system can test sequences of unknown composition to determine whether they contain any of the malicious signature fragments identified in the training process. A match between the test sequence fragment and a malicious signature fragment in the malicious signature database indicates that the test sequence may contain sequence information about a malicious organism, and the test sequence may be flagged for further scrutiny. Metadata stored about the malicious signature fragments that match a region of the test sequence fragment may be referenced to identify or classify the test sequence or test sequence fragment. For example, if multiple malicious signature fragments are found in a test sequence, common characteristics of the matching malicious signature fragments may be determined from the metadata. For example, it may be determined that all of the matching malicious signature fragments are from a particular sample (or related samples) of a particular organism, which may suggest that the customer is attempting to replicate that organism.
[0041] According to another approach, multiple signature databases may be used, with each signature database containing signature fragments for a particular known type or class of organism. For example, an influenza signature database may store signature fragments that are uniquely present in one or more sequences of influenza organisms, and similarly for an anthrax signature database. During the training process, the fragments in each signature database may be compared to one or more benign fragment databases, as in the approach described above, to filter out any fragments present in the benign fragment database, leaving only fragments for the organism represented by that particular signature database.
[0042] A test sequence of unknown composition can then be broken down into test sequence fragments and compared to each of a plurality of signature databases. The presence of a test sequence fragment in a particular signature database can indicate that the test sequence fragment contains information about a corresponding organism type. For example, a match between a test sequence fragment and a signature fragment in an influenza signature database can indicate that the test sequence contains part or all of a sequence for an influenza pathogen. Different test sequence fragments from a particular test sequence can match signature fragments in multiple signature databases. The number of matches and / or the location of the matches in the test sequence can be used to classify the test sequence or a region thereof according to one or more organism types for which it may contain sequence information.
[0043] It should be understood that the embodiments of the methods and apparatus described herein are not limited in application to the details of the configuration and arrangement of components set forth in the following description or illustrated in the accompanying drawings. The methods and apparatus can be implemented in other embodiments and can be practiced or performed in various ways. Examples of specific implementations are provided herein for illustrative purposes only and are not intended to be limiting. Additionally, the phrases and terms used herein are for descriptive purposes and should not be understood as limiting. The use of "including," "having," "having," "including," "involving," and variations thereof herein are meant to include the listed items and their equivalents as well as additional items. References to "or" can be construed as inclusive, such that a term described using "or" can refer to a single, a plurality, and all of the described terms. References to front, back, left, right, top, bottom, up, down, vertical, and horizontal are intended for convenience of description and do not limit the present systems and methods or components thereof to any one location or spatial orientation.
[0044] FIGURE 1 is a block diagram of a system 100 configured to perform a method for identifying target signature fragments, which may be malicious signature fragments. What is considered a "target organism" and what distinguishes a target organism from a "non-target organism" may be controlled or selected by a user. For clarity of explanation, FIGURE 1 provides an example in which a target organism may be a malicious organism and a non-target organism may be a benign organism, but it should be understood that this example is not limiting.
[0045] The system 100 includes a benign fragment database 110 configured to store a number of benign fragments 112, 114 derived from benign organism sequences (not shown). In some examples, the benign fragment database 110 can be considered a "non-target fragment database" since it stores fragments derived from non-target organism sequences. The system 100 further includes a candidate signature database 120 configured to store a number of candidate signature fragments 122, 124 derived from malicious organism sequences (not shown) or "target organism sequences", as well as metadata 122', 124' regarding the candidate signature fragments 122, 124. For example, the metadata 122', 124' may indicate taxonomic information indicative of the biological classification of the organism from which the candidate signature fragments 122, 124 are derived. It should be understood that the system 100 is described as including a number of different databases 110, 120, 140, 150 for purposes of illustration. In some examples, the system 100 may include additional or fewer databases (including a single database) configured to store the information stored by the databases 110 , 120 , 140 , 150 .
[0046] The system 100 also includes a processor 130 configured to compare each of the candidate signature fragments 122, 124 to the benign fragment database 110 to determine whether a given candidate signature fragment 122, 124 matches any of the benign fragments 112, 114. In at least one example, a given candidate signature fragment 122, 124 "matches" any of the benign fragments 112, 114 if the given candidate signature fragment 122, 124 is an exact match of any of the benign fragments 112, 114. If the candidate signature fragment 122 matches the benign fragment 112, it is known that the candidate signature fragment 122 does not uniquely identify a malicious organism sequence. On the other hand, if the candidate signature fragment 124 does not match any of the benign fragments 112, 114, the candidate signature fragment 124 may uniquely identify a malicious organism sequence. In that case, the candidate signature fragment 124 may be stored in the malicious signature database 140 as one of the malicious signature fragments 142, 144. In some examples, the malicious signature database 140 may be considered a "target fragment database" since it stores fragments derived from target organism sequences. The malicious signature database 140 may further store metadata 142', 144' associated with the malicious signature fragments 142, 144. For example, the metadata 142', 144' may indicate taxonomic information indicating the biological classification of the organism from which the malicious signature fragment 142, 144 is derived.
[0047] The system 100 further includes a test sequence database 150 configured to store a number of test sequences 152, 154. During a test operation of the system 100, one or more of the test sequences 152, 154 in the test sequence database 150 are compared to the malicious signature fragments 142, 144 to determine whether the malicious signature fragments 142, 144 are present in the one or more of the test sequences 152, 154. If so, any of the test sequences 152, 154 that match any of the malicious signature fragments 142, 144 may be flagged as including a sequence (or a signature fragment thereof) of a malicious organism (or more generally, a target organism). In some embodiments, the one or more test sequences 152, 154 may be a complete gene sequence (e.g., representing a complete DNA strand, other nucleic acid sequence, amino acid sequence, etc.). In other embodiments, the one or more test sequences 152, 154 may be a partial sequence of a given length. Here, an optimal length for testing is selected. In a preferred embodiment, the one or more test sequences 152, 154 are analyzed in their entirety, for example in sequential order. In some embodiments, the one or more test sequences 152, 154 may first be compared to the malicious signature fragments 142, 144 at positions on the one or more test sequences 152, 154 where a malicious signature is expected to be found. If no match is found, less likely positions may be examined.
[0048] In some embodiments, a user interface may be used to display or otherwise provide results of the comparison and / or to issue a warning or other communication that the test sequence may represent a malicious organism.
[0049] The benign fragment database 110 may be configured as a space-efficient, probabilistic data structure, such as a Bloom filter. Such a filter can be used to quickly and efficiently test whether an element is a member of a set. In the present context, such a filter can be used to quickly determine whether a candidate signature fragment matches (e.g., closely matches) one or more benign fragments in the benign fragment database 110 (in which case the candidate signature fragment is certainly not suitable as a malicious signature fragment), or alternatively, whether the candidate signature fragment does not match (e.g., does not closely match) any benign fragment in the benign fragment database 110 (in which case the candidate signature fragment may be suitable as a malicious signature fragment).
[0050] A "false positive" may occur when a candidate signature fragment does not match any benign fragment in the benign fragment database 110, but is nevertheless not unique to the sequence of the malicious organism. For example, a candidate signature fragment may match a benign fragment that would have been generated from a sequence of a benign organism from which no extraction had been performed. In this situation, a false positive identification of a malicious signature fragment may result in a sequence of a benign organism being erroneously identified as a sequence of a malicious organism during the testing phase of the operation, thereby necessitating additional (unnecessary) investigation. A sufficiently large number of benign fragments may be populated in the benign fragment database 110 to reduce the occurrence of false positives to an acceptable level; as the size of the benign fragment database 110 increases, the rate of false positives approaches zero. In one example for a given organism type, generating benign fragments from a set of 1.5 million base pairs of benign sequences may result in a false positive rate of 4%. Increasing the number of base pairs by a factor of 10 (to 11.5 million) would reduce the false positive rate to 0.25%.
[0051] In some embodiments, the benign fragment database 110 may be populated with benign fragments 112, 114 (e.g., from an external source) such that extraction by the system 100 of benign fragments 112, 114 from sequences of benign organisms is not required. In other embodiments, the benign fragment database 110 and / or the processor 130 may be configured to extract benign fragments 112, 114 from sequences obtained from one or more known benign organisms. The length n of the fragments may be configurable, and fragments of a given length n are referred to herein as n-grams. The examples shown here use 3-gram fragments, but any practical length n can be used.
[0052] FIG. 2A illustrates how an exemplary benign DNA sequence 202 ("CAGGTT") containing six nucleotides from a benign organism may be extracted into multiple 3-gram fragments 202a-d for storage in the benign fragment database 110, according to some embodiments. It should be understood that DNA is provided only as an example for purposes of illustration, and the principles of the present disclosure are applicable to other nucleic acids, such as RNA, amino acid sequences, and the like. Each of the 3-gram fragments 202a-d represents a subsequence of nucleotides of the benign DNA sequence 202 at a different starting point. For example, the first 3-gram fragment 202a includes a 3-nucleotide subsequence ("CAG") starting at a first position of the benign DNA sequence 202, the second 3-gram fragment 202b includes a 3-nucleotide subsequence ("AGG") starting at a second position of the sequence 202, and so on. Thus, a DNA sequence of length m may give rise to mn fragments.
[0053] Returning to Figure 1, the candidate signature database 120 and / or the processor 130 may be configured to derive the candidate signature fragments 122, 124 from sequences obtained from one or more known malicious organisms to be detected during a testing phase of operation of the system 100. The length of the candidate signature fragments 122, 124 may be selected to be the same as the n-gram benign fragments (e.g., 3).
[0054] 2B illustrates how a malicious DNA sequence 204 is extracted into multiple 3-gram fragments 204a-d for storage in the candidate signature database 120, according to some embodiments. The extraction is performed in much the same manner as for the benign DNA sequence 202 of FIG. 2A. For example, a first 3-gram fragment 204a includes a 3-nucleotide subsequence ("GCA") beginning at a first position of the malicious DNA sequence 202, a second 3-gram fragment 204b includes a 3-nucleotide subsequence ("CAG") beginning at a second position of the sequence 204, and so on.
[0055] The processor 130 is further configured to identify candidate signature fragments 122, 124 in the candidate signature database 120 that do not match any of the benign fragments 112, 114 in the benign fragment database 110. Such unmatched candidate signature fragments 122, 124 may be identified as malicious signature fragments 142, 144 and stored in the malicious signature database 140. With reference to Figures 2A and 2B, for example, candidate signature fragment 204b ("CAG") matches benign fragment 202a, candidate signature fragment 204c ("AGG") matches benign fragment 202b, and candidate signature fragment 204d ("GGT") matches benign fragment 202c. Thus, none of candidate signature fragments 204b, 204c, or 204d are identified as malicious signature fragments. However, candidate signature fragment 204a ("GCA") does not have a match in the benign fragment database 110 and is identified as a malicious signature fragment.
[0056] Returning to FIG. 1, the malicious signature fragments 142 , 144 identified by the processor 130 are stored in a malicious signature database 140 .
[0057] Each of the benign fragment database 110, the candidate signature database 120, and / or the malicious signature database 140 may be populated, populated, or optimized to improve performance. For example, duplicate fragments within a given database may be removed, and the fragments stored therein may be sorted or filtered for optimization purposes. As discussed in more detail below, for example, the target fragments may be filtered to identify a subset of universal target signatures. In some embodiments, the benign fragment database 110, the candidate signature database 120, and / or the malicious signature database 140 may be stored in an encrypted format or otherwise protected against access by unauthorized parties, and may be decrypted at or shortly before run-time.
[0058] The candidate signature database 120 and / or the malicious signature database 140 may also store metadata 122', 124', 142', 144' about the fragments they store or about the corresponding malicious organisms. Such metadata may include, for example, the date and time the fragment was created; an identifier for the sample / organism from which the fragment was obtained; the location of the fragment within that sample; a unique identifier for the fragment; the species or genus of the corresponding organism; the general category of the organism (e.g., virus, bacteria), etc.
[0059] FIG. 3 is a flow diagram of an example of a method 300 for identifying regions of malicious gene sequences.
[0060] The method 300 begins at step 310 .
[0061] In step 320, a plurality of benign fragments are identified. The plurality of benign fragments are derived from a first sequence obtained from at least one benign organism. As will be appreciated in light of the above, step 320 may more broadly include identifying a plurality of non-target fragments derived from a first sequence obtained from at least one non-target organism. In some embodiments, the plurality of benign fragments are extracted from sequences from one or more known benign organisms, as described above with reference to FIG. 2A. In particular, benign fragments of length n may be extracted from such benign sequences. The length n of such fragments may be selected to be sufficient to facilitate the identification of malicious signature fragments at a later stage. For example, consider a scenario in which fragments of length n=20 are required to uniquely identify a malicious sequence. In other words, there will be no fragments of length n<20 that are present in the malicious sequence but not present in the benign sequence. In such a scenario, the length n of both the benign fragments and the malicious signature fragments may be 20 (or more).
[0062] As described above with reference to FIG. 2A, benign fragments may be extracted from a benign sequence by creating n-length fragments, if possible, that correspond to subsequences starting at each position in the benign sequence. In other words, each position (e.g., nucleotide) in the benign sequence and the following n positions (if any) may be represented by a benign fragment. In other embodiments, only certain positions in the benign sequence may be used as the basis for benign fragments. For example, due to the chemical nature of the sequence or other parameters, it may be impossible or unlikely that a bad signature fragment will correspond to certain starting positions on the benign sequence. In that case, those specific starting positions may not be used as the basis for extracting benign fragments.
[0063] In some embodiments, the extraction of benign fragments from one or more benign sequences may not be performed by the system. Rather, the extraction may already be performed and the benign fragments may be provided to the system, for example, by a third party. For example, a database of benign fragments may be made available. In another example, the benign fragments may have been extracted and maintained by the system during a previous operation, and thus do not need to be extracted again. It will be appreciated that the extraction and / or use of benign fragments to train the system may be performed in a rolling manner, i.e., new benign fragments may be added over time to improve the accuracy of the results.
[0064] The plurality of benign fragments may be derived, in whole or in part, from at least one benign organism having at least one characteristic relevant or useful for identifying a malicious signature fragment. In some embodiments, benign organisms similar in some way to the malicious organism of interest may be used to extract the benign fragments. The similarity between the benign and malicious organisms may reflect the similarity of their genetic sequences, such that the system can identify a relatively small number of differences as fragments of the malicious signature. For example, the benign organism may be a non-malicious strain of the malicious microorganism of interest. In another example, the benign organism and the malicious organism may belong to a common genus or a broader range of related organisms. In an example where the target organism includes SARS-CoV-2 and its malicious variants, the benign organism may include other coronaviruses, such as benign coronaviruses and / or coronaviruses that are less virulent than SARS-CoV-2. In this example, a user can determine which organisms (e.g., which coronaviruses) are considered benign, and it is from these selected organisms that the plurality of benign fragments are derived. As discussed in more detail below, certain organisms may alternatively be classified as "neutral" rather than as target or non-target organisms.
[0065] In step 330, a plurality of candidate signature fragments are extracted from a second sequence obtained from a malicious organism. As will be understood in light of the above, step 330 may more broadly include identifying a plurality of candidate signature fragments derived from a second sequence obtained from a target organism. Step 330 may be performed repeatedly to extract a plurality of candidate signature fragments from respective sequences obtained from each of a plurality of malicious (or target) organisms. In some embodiments, the extraction is performed as described above with reference to FIG. 2B. In particular, candidate signature fragments of length n may be extracted from sequences obtained from one or more malicious organisms. The length n of such fragments may be selected to be sufficient to facilitate identification of malicious signature fragments at a later stage.
[0066] At stage 340, for each of the plurality of candidate signature fragments, it is determined whether the candidate signature fragment matches at least one of the plurality of benign fragments. In at least one example, stage 340 includes determining whether the candidate signature fragment exactly matches at least one of the plurality of benign fragments. In some embodiments, the plurality of benign fragments are arranged in a benign fragment database as a probabilistic data structure (e.g., a Bloom filter), and for each candidate signature fragment, a query is performed on the Bloom filter. In other embodiments, the plurality of benign fragments are stored as an array, a search tree, a relational database, a schema-free database, a collection of n-tuples, or in other manners and queried appropriately. In some embodiments, the plurality of benign fragments are deduplicated, sorted, and / or filtered to increase efficiency.
[0067] At step 350, a candidate signature fragment is identified as a malicious signature fragment in response to the candidate signature fragment not matching the at least one of the plurality of benign sequence fragments. In some embodiments, a candidate signature fragment identified as a malicious signature fragment may be stored as a malicious signature fragment in a malicious signature database along with any metadata about the malicious signature fragment. In other embodiments, a separate malicious signature database may not be used, and candidate signature fragments may be flagged as malicious signature fragments in the candidate signature database, and at the end of the training process, candidate signature fragments that are not flagged as malicious signature fragments may be discarded or otherwise not used during the testing process.
[0068] The process 300 ends at step 360 .
[0069] In some examples, once the training process is complete, the test sequence may be examined to determine whether one or more bad signature fragments are present. If present, the test sequence may be flagged for further scrutiny and / or considered for rejection from the synthesis / replication process in examples where synthesis / replication is performed, e.g., as described below with respect to FIG. 4. In various examples, once process 300 is complete, the bad signature fragments (or "target signature fragments") are further processed. For example, the target signature fragments may be filtered to identify one or more universal target signature fragments and / or one or more low homology universal signature fragments, as described below with respect to FIG. 10. Additionally, in some examples, the test sequence may be examined to determine whether one or more universal target signature fragments and / or one or more low homology universal signature fragments are present in the test sequence. That is, in some examples, the malicious signature fragments provided in process 300 may be filtered according to FIG. 10 to identify universal target signature fragments, and the universal target signature fragments may be used to test test sequences according to FIG. 4.
[0070] FIG. 4 is a flow diagram for one example of a method 400 for testing one or more test sequences for the presence of malicious signature fragments.
[0071] The method 400 begins at step 410 .
[0072] At step 420, it is determined whether a bad signature fragment is present in at least one test sequence. In one example, the bad signature fragment may include a candidate signature fragment identified as a bad signature fragment at step 350 described above. In another example, the bad signature fragment may include a universal target signature fragment identified at step 1008 described below. In another example, the bad signature fragment may include a low homology universal target signature fragment identified at step 1010 described below.
[0073] In some embodiments, the at least one test sequence is a sequence provided for the purpose of replication. The sequence may represent a single gene sequence or may include regions that are intended to be "cut and spliced together" later using mechanisms such as CRISPR. In some embodiments, the at least one test sequence may be a complete genetic sequence (e.g., representing a complete DNA strand). In other embodiments, the at least one test sequence may be a subsequence of a complete gene sequence. The optimal length of the subsequence, or the portion of the complete gene sequence that is included in the subsequence, may be selected. For example, the test sequence may be a subsequence of a complete gene sequence, with the subsequence selected from a position or region of the complete gene sequence based on the likelihood of finding a malicious signature fragment within that region. In yet other embodiments, the subsequence may be selected to omit a known benign region of the complete gene sequence. In some examples, the at least one test sequence may be analyzed in real time or near real time as the at least one test sequence is received, as described below with respect to FIG. 11. For example, the at least one test sequence may be analyzed in real time as a sequencer generates the at least one test sequence.
[0074] At each sequential position on the at least one test sequence, a bad signature fragment may be compared to the at least one test sequence. For example, a 3-gram bad signature fragment may be first compared to positions 1-3 on the at least one test sequence, then to positions 2-4 on the at least one test sequence, etc.
[0075] In some embodiments, the number and type of matches for each of the at least one test sequence and / or malicious signature fragments may be stored. For example, data may be stored indicating the location of each malicious signature fragment on the at least one test sequence, the type of malicious signature fragment, the number of times each malicious signature fragment occurs in the at least one test sequence, and other information.
[0076] The malicious signature fragments and / or metadata about the corresponding malicious organisms may be used to identify or classify the test sequence. For example, if multiple malicious signature fragments are found in the test sequence, common features of the matching malicious signature fragments may be determined from the metadata. For example, it may be determined that all the matching malicious signature fragments are from a particular sample (or related samples) of a particular organism, which may suggest that the customer is attempting to replicate that organism. Depending on the number of signature fragments in the test sequence, it may be possible to identify the genus, species, or even the particular sample of the malicious organism reflected in the test sequence.
[0077] The type and number of signature fragments corresponding to malicious organisms or types of organisms can be tracked and analyzed to draw conclusions about the test sequence. For example, if the number, cumulative length, or other statistics of influenza signature fragments in a test sequence exceed a given threshold, it can be automatically concluded that the test sequence is an attempt to synthesize influenza. In another embodiment, such statistics can be used to determine a level of confidence in determining that a sequence has been submitted for nefarious purposes.
[0078] In optional step 430, a decision may be made for the at least one test sequence. For example, depending on the number and type of malicious signature fragments occurring on the at least one test sequence and the malicious organisms they are associated with, a decision may be made to reject the at least one test sequence from a synthesis / replication application in non-limiting examples including synthesis / replication applications, and / or to flag the at least one test sequence for further review by the system and / or user. In some embodiments, a threshold number of occurrences of malicious signature fragments may be set, and a decision may be made for the at least one test sequence based on whether the threshold is exceeded. Different thresholds may be set for different malicious organisms, with more dangerous pathogens having a low / zero threshold and less dangerous pathogens having a higher threshold.
[0079] The method 400 ends at step 440 .
[0080] In addition to the above-described approaches for identifying the presence of sequences of malicious organisms, there are also applications in which it is useful to rapidly classify a test sequence as one or more of a number of organisms (including but not limited to pathogens).
[0081] 5 is a block diagram for a system 500 configured to perform a method for identifying signature fragments that can be used to classify an unknown sequence. System 500 may be similar to system 100 in some aspects, with some differences being discussed herein.
[0082] The system 500 includes at least one benign fragment database 510 configured to store a number of benign fragments 512, 514 derived from sequences of benign organisms (not shown). The system 500 further includes a number of malicious signature databases 520a-520c configured to store a number of candidate signature fragments 522a-522c, 524a-524c derived from sequences of malicious organisms (not shown). In this approach, each of the malicious signature databases 520a-520c may be organized as a probabilistic data structure, such as a Bloom filter. Each of the malicious signature databases 520a-520c may correspond to a different type or group of malicious organisms. For example, malicious signature database 520a may store signature fragments for the influenza organism, malicious signature database 520b may store signature fragments for the anthrax organism, and malicious signature database 520c may store signature fragments for the smallpox virus. Each malicious signature database may also store metadata (not shown) about the fragments stored therein, as described above with respect to malicious signature database 140.
[0083] The system 500 further includes a processor 530 configured to compare the candidate signature fragment (not shown) to a plurality of benign fragment sequences 512, 514. If no match is found, the particular candidate signature fragment may be determined to be a suitable signature fragment for a particular type of malicious organism associated with one of the malicious signature fragment databases 520a-520c. If so, the candidate signature fragment may be stored in one of the malicious signature fragment databases (e.g., 520b) corresponding to the type of malicious organism. Continuing with the previous example, if the candidate signature fragment is found to be a suitable signature fragment for influenza, the candidate signature fragment may be stored as signature fragment 522a in the malicious signature fragment database 520a.
[0084] As in system 100, the candidate signature fragments in system 500 may be extracted from sequences of known organisms. While the examples discussed herein involve malicious organisms, it will be understood that the same techniques may be used to identify or classify non-malicious organisms of interest. The candidate signature fragments may be stored in one or more candidate signature fragment databases (not shown).
[0085] If candidate signature fragments are stored for several malicious organisms or types of malicious organisms, the candidate signature fragments may be stored in one or more databases in any number of ways that allow the candidate signature fragments to be associated with a particular organism or type of organism. In one embodiment, the candidate signature fragments may be stored in a single candidate signature fragment database, where each candidate signature fragment is associated (by an identifier or other association) with a particular malicious organism or type of malicious organism. In other embodiments, the candidate signature fragments may be stored in different databases according to their associated malicious organisms or types of organisms.
[0086] The benign fragments may likewise be stored in a common database or may be stored separately according to the type of benign organism from which they originate, or according to the type of malignant organism or organism for which they are used to identify the signature fragment.
[0087] Similar to system 100, system 500 further includes a test sequence database 550 configured to store a number of test sequences 552, 554. During a test operation of system 500, one or more of the test sequences 552, 554 in test sequence database 550 are compared to the signature fragments in one or more of malicious signature fragment databases 520a-520c to determine whether any of the malicious signature fragments 522a-c, 524a-c are present in the one or more of test sequences 552, 554. For example, the test sequences 552, 554 may be applied to the Bloom filters of each of the malicious signature fragment databases 520a-c to determine whether any matches are found. If found, any of the test sequences 552, 554 that match any of the malicious signature fragments 522a-c, 524a-c may be flagged as containing a sequence (or a fragment thereof) of a malicious organism associated with the malicious signature fragment database 520a-c that contains such malicious signature fragment.
[0088] In some embodiments, the one or more test sequences 552, 554 may be complete gene sequences (e.g., representing a complete DNA strand). In other embodiments, the one or more test sequences 552, 554 may be subsequences of a given length, where an optimal length for testing is selected. In some embodiments, the one or more test sequences 552, 554 may first be compared to the malicious signature fragments 522a-c, 524a-c at positions on the one or more test sequences 552, 554 where one would expect to find a malicious signature. If no match is found, less likely positions may be examined.
[0089] FIG. 6 is a flow diagram of an example of a method 600 for classifying regions of a genetic sequence using a system such as 500.
[0090] The method 600 begins at step 610 .
[0091] In step 620, a first plurality of sequence fragments is generated from a first plurality of organisms having a first trait, and in step 630, a second plurality of sequence fragments is generated from a second plurality of organisms having a second trait. In some embodiments, the extraction is performed as described above with reference to FIG. 2B. Specifically, candidate signature fragments of length n may be extracted from sequences obtained from a plurality of malicious organisms having the first trait and from a plurality of malicious organisms having the second trait. A trait may be a classification or category of a type of organism (e.g., influenza, coronavirus, etc.).
[0092] In step 640, a plurality of benign sequence fragments are identified. Step 640 may be performed in much the same manner as step 320 of method 300. As noted above, in some embodiments, the extraction of the plurality of benign sequence fragments may not be performed by the system. Rather, the extraction may have already been performed and the benign sequence fragments may be provided to the system, for example, by a third party. For example, a database of benign fragments may be made available. In another example, the benign sequence fragments may have been extracted and maintained by the system during a previous operation.
[0093] In step 650, the first plurality of candidate sequence fragments is filtered to remove at least one of the plurality of benign sequence fragments, and in step 660, the second plurality of candidate sequence fragments is filtered to remove at least one of the plurality of benign sequence fragments. Other plurality of candidate sequence fragments may also be filtered. The method is not limited to two such plurality. Steps 650 and 660 may be performed in a manner similar to step 340 of method 300. In particular, each plurality of candidate signature fragments may be compared to benign fragments (e.g., in a Bloom filter), and any candidate signature fragments that match the benign fragments may be identified as not being suitable signature fragments. Signature fragments found to be suitable may be stored in one of the malicious signature fragment databases corresponding to the type of organism uniquely identified by the signature fragment. The malicious signature fragment database may be organized as a plurality of probabilistic data structures, such as Bloom filters.
[0094] The method 600 ends at step 670 .
[0095] In some examples, once the training process is complete, the test sequence may be examined to determine whether one or more bad signature fragments are present; if present, the test sequence may be flagged for further scrutiny and / or considered for rejection from the synthesis / replication process in non-limiting examples including, for example, the synthesis / replication process, as described below with respect to FIG. 7. In various examples, once method 600 is complete, the bad signature fragments (or "target signature fragments") are further processed. For example, the target signature fragments may be filtered to identify one or more universal target signature fragments and / or one or more low homology universal signature fragments, as described below with respect to FIG. 10. Further, in some examples, a test sequence may be examined to determine whether one or more universal target signature fragments and / or one or more low homology universal signature fragments are present in the test sequence; i.e., in some examples, the bad signature fragments provided in method 600 may be filtered according to FIG. 10 to determine universal target signature fragments, and the universal target signature fragments may be used to examine the test sequence according to FIG. 7.
[0096] FIG. 7 is a flow diagram of an example method 700 for testing one or more test sequences for the presence of malicious signature fragments stored in one or more malicious signature databases.
[0097] The method 700 begins at step 710 .
[0098] In step 720, it is determined whether a malicious signature fragment is present in the at least one test sequence. Step 720 may be performed similarly to step 420 of method 400. The at least one test sequence or fragments thereof may be compared to the one or more signature fragments stored in a plurality of fragment signature fragment databases (e.g., in a Bloom filter). In some embodiments, the test sequence may be compared to all signature fragment databases, or some standard subset of the signature fragment databases. In other embodiments, a particular signature fragment database may be selected for comparison based on some known characteristic of the test sequence. For example, if it is determined that the test sequence is more likely to contain a gene sequence for a coronavirus, the test sequence may not be compared to a signature fragment database unrelated to coronaviruses. In some embodiments, the signature fragment database to which the test sequence is compared may be selectable by a user, e.g., an operator of system 500.
[0099] A determination may be made for the at least one test sequence in optional step 730. Step 730 may be performed similarly to step 430 of method 400.
[0100] The method 700 ends at step 740 .
[0101] As described above, process 300 may be performed to identify target signature fragments at step 350. Similarly, process 600 may be performed to identify target signature fragments at steps 650 and 660. Processes 300, 600, or each of their individual steps, may be performed several times to identify several target signature fragments. In some examples, several target signature fragments may be identified from a single organism (e.g., a single SARS-CoV-2 virus). In other examples, several target signature fragments may be identified from each of several organisms (e.g., including several coronaviruses) in performing either or both of processes 300, 600.
[0102] A group of organisms may include target organisms and non-target organisms. For example, consider the group of coronaviruses. Some coronaviruses may be considered non-target viruses, at least because they are not harmful or minimally harmful to humans. Other coronaviruses, such as the SARS-CoV-2 coronavirus, may be considered targeted, at least because they are harmful to humans. Still other coronaviruses, such as the SARS-CoV-1 coronavirus, may be considered neutral, at least because they may be harmful to humans but have been nearly eradicated from the world. Thus, such coronaviruses may be identified as neutral, since the likelihood of such coronaviruses being found in a sample may be extremely low and therefore may be considered unimportant.
[0103] It may be advantageous to be able to distinguish certain target coronaviruses from other non-target or neutral coronaviruses in a sample in order to determine whether the coronaviruses in the sample are virulent or benign so that appropriate action can be taken. Thus, it may be advantageous to define a first group of coronaviruses as target coronaviruses (e.g., SARS-CoV-2 virus and its genetic variants), a second group of coronaviruses as non-target coronaviruses, and a third optional group of coronaviruses as neutral coronaviruses. Furthermore, it may be advantageous to identify one or more universal target signature fragments that universally identify all coronaviruses in the first group of coronaviruses but not any coronaviruses in the second group of coronaviruses. Because the third group of coronaviruses is a neutral group, it may not be important whether the universal target signature fragment identifies any coronaviruses in the third group.
[0104] For example, FIG. 9 shows a diagram 900 of several sets of target signature fragments identified from several target organisms (e.g., the first group of coronaviruses discussed above), according to one example. The diagram 900 includes a first collection of target signature fragments 902, a second collection of target signature fragments 904, and one or more universal target fragments 906 that represent the mathematical intersection of the two collections of fragments 902, 904. Although the diagram 900 includes two collections of target signature fragments for illustrative purposes, in other examples, the principles discussed herein may be applicable to any other multiple collections of target signature fragments. Continuing with the above example, if the first group of target coronaviruses includes two coronaviruses, the two collections of target signature fragments 902, 904 may be considered as collections of target signature fragments for each of those two coronaviruses.
[0105] The first set of target signature fragments 902 may be identified from a first organism (such as a SARS-CoV-2 virus). For example, the first set of target signature fragments 902 may be identified in step 350 of method 300, or steps 650 and / or 660 of method 600, by analyzing a sequence of the first organism. The second set of target signature fragments 904 may be identified from a second organism (such as a genetic variant of the SARS-CoV-2 virus). For example, the first set of target signature fragments 902 may be identified during a subsequent performance of step 350 of process 300, or steps 650 and / or 660 of process 600, by analyzing a sequence of the second organism.
[0106] One or more target signature fragments may be in both the first set of target signature fragments 902 and the second set of target signature fragments 904. In Figure 9, these universal target signature fragments 906 are represented by the mathematical intersection of the sets 902, 904. The universal target signature fragments 906 are "universal" because they are present in all sets 902, 904 in the drawing 900.
[0107] In some examples, a set of universal target signature fragments may be identified for a group of organisms, such as a group of target coronaviruses. The set of universal target signature fragments may be significantly smaller than the combination of all sets of target signature fragments from which the universal target signature fragments are derived. For example, in FIG. 9, to the extent that the sets 902, 904 contain at least one uncommon target signature fragment, the size of the universal target signature fragment 906 will be smaller than the combination of the size of the first set of target signature fragments 902 and the second set of target signature fragments 904, since the at least one uncommon target signature fragment is excluded from the universal target signature fragment 906. It may be advantageous to identify a set of universal target signature fragments and use the set of universal target signature fragments to determine whether the universal target signature fragment is present in at least one test sequence (e.g., as described above in steps 420 and 720). In one non-limiting example, a set of universal target signature fragments may be identified from a set of target signature fragments derived from similar organisms. For example, a set of target signature fragments may be derived from the SARS-CoV-2 virus and from each of several genetic variants of the SARS-CoV-2 virus. A set of universal target signature fragments may be identified from these sets. The resulting set of universal target signature fragments may be applied to one or more test sequences to, for example, predict whether a given test sequence is derived from the SARS-CoV-2 virus or a genetic variant thereof.
[0108] This may be advantageous at least because, as described above, the size of the set of universal target signature fragments may be significantly smaller than the size of the combination of the sets of target signature fragments used to create the set of universal target signature fragments. Thus, analyzing the at least one test sequence may be significantly faster, at least because there are fewer target signature fragments to compare the at least one test sequence with. Furthermore, the number of false positives may be reduced, at least because if a target signature fragment is present in all sets of target signature fragments, it is less likely to be spurious. In contrast, a target signature fragment that appears in the sequence of only one target organism is more likely to be spurious and to generate a false positive. In addition to identifying a set of universal signature fragments, it may be advantageous to filter the set to identify only universal signatures that evidence low homology with respect to background sequences, in order to further reduce false positives.
[0109] The universal signatures, including low homology universal signatures, may be used in various implementations. For example, the universal signatures may be used as physical probes in analyzing samples for the presence of genes of interest. For example, the universal signatures may be used as primers in a polymerase chain reaction (PCR) process to "instruct" the polymerase on which genes (e.g., certain coronavirus genes) to bind to. A PCR process may then be performed to amplify the genes of interest represented by the universal signature in the sample. In various examples, the universal signatures advantageously amplify the genes of interest without amplifying other genes, such as genes from pathogens that are not the pathogen of interest. For example, if the pathogen of interest is the SARS-CoV-2 coronavirus, the universal signatures may amplify the SARS-CoV-2 coronavirus without amplifying other coronaviruses that are not of interest. The samples containing the amplified genes of interest may then be used in subsequent analyses.
[0110] 10 illustrates a process 1000 for filtering signature fragments, according to an example. Process 1000 may be performed, for example, by processor 130, processor 530, a combination of both, or another computing device(s). For purposes of explanation, a description of process 1000 is provided with respect to processor 530 and other aspects of system 500.
[0111] In step 1002, the process 1000 begins.
[0112] In step 1004, the processor 530 identifies a group of target organisms for identifying a set of universal target signature fragments. In some examples, the group of target organisms may be specified by a user. Continuing with the above example, the group of target organisms may include a first group of coronaviruses, which may include virulent coronaviruses (not benign or neutral coronaviruses). For example, a user may identify the SARS-CoV-2 virus and its genetic variants as a group of target organisms. It should be understood that the target "organism" is designated for illustrative purposes only, and that the principles of the present disclosure are applicable more broadly and generally to "targets"; in some examples, the process 1000 may be performed with respect to synthetic sequences in addition to or instead of organism sequences. That is, the group of targets identified in act 1004 may include target organism sequences, target synthetic sequences, or a combination of both.
[0113] In step 1006, the processor 530 obtains a set of target signature fragments for each target organism in the group of target organisms. The group of target organisms may include organisms for which a set of target signature fragments has been identified and stored in the database. For example, any of the signature fragment databases 520a-520c may store sets of target signature fragments for the SARS-CoV-2 virus and its genetic variants. Thus, the processor 530 may obtain the sets of target signature fragments by requesting the sets of target signature fragments for each target organism from the corresponding one or more of the signature fragment databases 520a-520c. In another example, a set of target signature fragments may not yet be identified for at least one of the target organisms. For each such target organism, the processor 530 and / or the processor 130 may execute the process 300 and / or the process 600 to identify a set of target signature fragments for that target organism.
[0114] As described above with respect to process 300, 600, the target signature fragments identified for a given organism may vary based on which sequences are selected as background (e.g., which sequences are identified as benign in step 320). In various examples, in performing process 300, 600, processor 130, 530 may identify a particular group of organisms to act as background. Continuing with the above example, consider a collection of coronaviruses (e.g., all known coronaviruses). Each coronavirus in the collection of coronaviruses may be classified (e.g., by a user) as belonging to either a first group of targets (e.g., virulent coronaviruses such as SARS-CoV-2 and its genetic variants), a second group of non-targets (e.g., benign coronaviruses), or a third group of neutral ones (e.g., SARS-CoV-1).
[0115] Using process 300 as an example, the plurality of benign fragments identified in step 320 may be derived from a second group of non-targets. The plurality of candidate signature fragments identified in step 330 may be derived from a first group of targets. A third group of neutrals may not be used in performing process 300. Thus, the target signature fragments identified in step 350 may be able to classify a given coronavirus sample as belonging to a target group or a non-target group. The target signature fragments may or may not identify coronaviruses classified in the neutral group, although as noted above, it may not be important whether or not a coronavirus in the neutral group is identified.
[0116] In step 1008, processor 530 filters the sets of target signature fragments obtained in step 1006 to identify a set of universal target signature fragments. The universal target signature fragments may include target signature fragments present in each and every set of target signature fragments. In one example, processor 530 analyzes each target signature fragment in all sets of target signature fragments obtained in step 1006 to determine whether each target signature fragment is present in all sets of target signature fragments obtained in step 1006. Determining whether each target signature fragment is present in each set of target signature fragments may be substantially similar to process 400, where the "bad fragments" in step 420 are each target signature fragments and the "at least one test sequence" in step 420 includes all target signature fragments in all sets of target signature fragments obtained in step 1006. In another example, determining whether each target signature fragment is present in each set of target signature fragments may include running a matching algorithm, such as an element-wise matching algorithm that matches each target signature fragment against all other target signature fragments in all sets of target signature fragments obtained in step 1006. In yet other examples, other methods may be implemented to determine whether a target signature fragment is present in all sets of target signature fragments obtained in step 1006.
[0117] In some examples, if a target signature fragment is not present in all sets of target signature fragments obtained in operation 1006, the target signature fragment is not added to the set of universal target signature fragments. In other examples, if the target signature fragment is present in at least a threshold number of sets of target signature fragments obtained in step 1006, the target signature fragment is added to the set of universal target signature fragments. For example, the threshold number may be a particular number of sets of target signature fragments or a particular threshold percentage of the set of target signature fragments obtained in step 1006. The threshold amount may be static or variable. For example, the threshold number may change as the number of sets of target signature fragments obtained in step 1006 changes.
[0118] In some examples, operation 1008 may include determining whether the set of universal target signature fragments includes at least a threshold number of universal target signature fragments. If the set does not include at least a threshold number of universal target signature fragments (e.g., if there are no or insufficient universal target signature fragments), process 1000 may proceed to step 1014 and end. In other examples, a notification may be provided to the user, but process 1000 may continue upon the user providing instructions to the contrary. In yet other examples, the conditions for target signature fragments to be added to the set of universal target signatures may be relaxed automatically and / or as directed by the user until the set of universal target signatures includes a threshold number of fragments.
[0119] In step 1010, the processor 530 identifies low homology universal target signature fragments from the set of universal target signature fragments identified in step 1008. Homology may refer to a degree of similarity between a universal target signature fragment and one or more background sequences, such as a plurality of benign fragments identified in step 320 and / or stored in either or both of the benign fragment databases 110, 510. It may be advantageous to identify low homology universal target signature fragments from the set of universal target signature fragments, i.e., universal target signature fragments that are not very similar to the one or more background sequences, because such significantly unique target signature fragments are less likely to be false positives where the test sequence is slightly different from the background sequence.
[0120] The degree of homology may be determined by running an algorithm for determining the degree of similarity between biological sequences, for example, a basic local alignment search tool (BLAST) algorithm. The output of such an algorithm may be expressed as or used to determine a homology parameter, such as a percentage of homology between biological sequences, the number of shared primers, etc. Step 1010 may include the processor 530 identifying the universal target signature fragments that have a homology parameter that meets certain criteria, such as within a certain threshold. For example, a low homology criterion may include having more than 80% homology and not sharing one or more primers with the one or more background sequences. That is, in this example, if a universal target signature fragment shares more than 80% and primers with the one or more background sequences, the universal target signature fragment may be rejected as having too high a homology with respect to the one or more background sequences. Thus, in step 1010, the processor 530 filters the set of universal target signature fragments to identify a smaller (or equal sized) set of low homology universal target signature fragments.
[0121] In step 1012, the processor 530 outputs the set of low homology universal target signature fragments. In some examples, the processor 530 may also output the set of universal target signature fragments. For example, the set of low homology universal target signature fragments (and / or the set of universal target signature fragments) may be output to and stored in any of the databases 140, 520a-520c. The set may then be implemented in the analysis of one or more test sequences, for example, as described above with respect to processes 400, 700 (or as described below in process 1100). That is, the test sequences may be analyzed to determine whether universal target signature fragments and / or low homology universal target signature fragments are present in the test sequences, and if so, to determine that the test sequences may be derived from the target organism (or synthetic sequences).
[0122] In operation 1014, the process 1000 ends.
[0123] Thus, process 1000 allows multiple sets of target signature fragments for multiple target organisms to be condensed into a single, smaller set that may be implemented in place of the multiple sets of target signature fragments, such as in cases where it may be advantageous to perform a faster analysis with fewer false positives.
[0124] For illustrative purposes, a non-limiting example is provided. In the following example, process 1000 is performed to identify a set of low homology universal target signature fragments from multiple sets of target signature fragments. Each set of target signature fragments may be capable of identifying a respective target coronavirus, such as SARS-CoV-2 or its genetic variants. All sets together may be capable of identifying any known coronavirus classified as a target, distinct from co-non-target or neutral coronaviruses.
[0125] In step 1004, a group of target organisms is identified. As described above, this group may include target coronaviruses. These target coronaviruses may be virulent coronaviruses. For example, the target coronaviruses may include SARS-CoV-2 and its genetic variants.
[0126] In step 1006, target signature fragments are obtained for each target organism. At least one of these target signature fragments may have already been previously obtained and stored in a database (e.g., malicious signature database 140), in which case step 1006 includes obtaining these target signature fragments from the database. Additionally, at least one of these target signature fragments may not have been obtained yet, in which case a process for obtaining the target signature fragments (e.g., process 300 or 600) may be performed. Process 1000 may proceed to step 1008 once all target signature fragments have been obtained.
[0127] As described above, a particular background may be selected in identifying the target signature fragment. For example, if a particular coronavirus (e.g., SARS-CoV-2 and its genetic variants) is targeted, the background may be or may include all other coronaviruses that may be considered benign. In some instances, as described above, certain coronaviruses may be classified as neutral.
[0128] In step 1008, the target signature fragments are filtered to identify one or more universal target signature fragments. As described above, the universal target signature fragments can include fragments that are universal among the target signature fragments for all target organisms (or at least a certain threshold percentage of target organisms). For example, the universal target signature fragment can be universally present in all target coronaviruses and absent in all non-target coronaviruses. In instances where a neutral is identified, it may be irrelevant whether a universal target signature fragment is present in the neutral. Thus, in step 1008, a reduced number of high efficacy signatures can be identified for subsequent classification of test sequences.
[0129] In step 1010, low homology universal target signature fragments are identified from the universal target signature fragments. For example, a BLAST algorithm can be implemented to determine the degree of homology between the universal target signature fragments and background coronaviruses. This may be computationally more practical than determining the degree of homology between the target signature fragments obtained in step 1006 and background coronaviruses, at least because the size of the universal target signature fragments may be significantly smaller than the size of the combination of the target signature fragments.
[0130] If any universal target signature fragment is determined to exhibit a degree of homology that is sufficiently significant (e.g., exhibits at least 80% homology), the universal target signature fragment may be identified as not being a low homology universal target signature fragment because the universal target signature fragment is too similar to the background coronavirus. Such high homology fragments may be too similar to the background coronavirus and thus more likely to produce false positives. Conversely, if the universal target signature fragment exhibits a low degree of homology (e.g., exhibits less than 80% homology), the universal target signature fragment may be identified as a low homology universal signature fragment. At a minimum, low homology universal signature fragments may be advantageous because low homology universal target signature fragments may be less likely to produce false positives by virtue of their low degree of homology with the background coronavirus.
[0131] In step 1012, the low homology universal target signature fragments and / or the universal target signature fragments are output. For example, the processor 530 can output the fragments by providing the fragments to a database, such as the malicious signature database 140. As described above, the fragments may be used later to classify the test sequence. For example, the low homology universal target signature fragments or the universal target signature fragments can be used to analyze a test sequence of a coronavirus to determine whether the coronavirus is a target coronavirus or a non-target coronavirus. At least, because there are fewer fragments filtered and therefore fewer fragments to compare samples to, it may be possible to perform the analysis more quickly than if a larger number of target signature fragments were implemented (e.g., using the set of target signature fragments obtained in step 1006). Thus, the examples disclosed herein allow for relatively fast and efficient classification of test sequences.
[0132] The test sequence (or a portion thereof) to be analyzed for the presence or absence of a target can be received in several ways. In some examples, the complete test sequence or a portion thereof may be stored in a file, and the file may be transmitted to a computing device configured to access the test sequences in the file and classify the test sequences. The test sequence may be determined by a sequencer, such as a DNA sequencer. As will be appreciated by those skilled in the art, a sequencer is used to automate the gene sequencing process by analyzing a sample, determining the genetic sequence of the sample, and outputting a genetic readout that includes the genetic sequence of the sample. The readout may be output as a text string, which may be stored in a file, as described above.
[0133] In some examples, the principles of the present disclosure may be applied to the output of a sequencer before the sequencer has finished sequencing the sample, for example, genetic sequences may be analyzed (including, for example, classifying sequences as target or non-target) as the sequences are output from the sequencer in substantially real-time or near real-time.
[0134] FIG. 11 illustrates a process 1100 for analyzing the output of a sequencer, according to one example. In various examples, the process 1100 may be performed by a processor, such as the processor 130, 530. The processor may be a component of the sequencer in some examples. In other examples, the processor may be a component of a device other than the sequencer. For example, the processor may be a component of a device that can be communicatively coupled to the sequencer to receive the output of the sequencer in real time or near real time. In other examples, the processor may be a component of the other device.
[0135] At step 1102, the process 1100 begins.
[0136] In step 1104, a target group is identified. For example, a group of organisms to be searched for in the test sequence may be identified. In some examples, the target group may be selected by a user. For example, the user may select a certain group of coronaviruses, such as SARS-CoV-2 and its genetic variants.
[0137] In step 1106, a target signature fragment is obtained for each target identified in step 1104. The target signature fragment may include a universal target signature fragment and / or a low homology universal target signature fragment. In some examples, the target signature fragment may include a fragment identified in steps 350, 650, and 660, for example, described above. The target signature fragment may be obtained by accessing a database that includes target signature fragments, such as one of databases 140, 520a-520c.
[0138] In step 1108, a portion of the test sequence is received. The portion of the test sequence may be received in real time or near real time from the output of the sequencer.
[0139] Step 1108 may be performed before the sequencer has finished sequencing a particular sample. Thus, it should be appreciated that the portion of test sequences received in step 1108 may only be a portion of the total gene sequence of the sample that may ultimately be provided by the sequencer as a whole. Additionally, the sequencer may be analyzing a sample that includes multiple organisms, each having a respective gene sequence. Thus, step 1108 may include receiving sequence fragments of multiple gene sequences.
[0140] In step 1110, the portion of the test sequence received in step 1108 is filtered. Step 1110 may be performed in parallel with step 1108. That is, a certain portion of the test sequence may be filtered in step 1110, while an unfiltered portion of the test sequence may be received in step 1108. Filtering the portion of the test sequence may include removing low quality reads from the portion of the test sequence. Low quality reads may include sequence information that is likely to be inaccurate, i.e., information that is unlikely to represent the true gene sequence of the sample. Such inaccuracies may be introduced by the sequencer mis-sequencing the sample. It may be advantageous to filter out low quality reads, at least because the reads may contain errors and thus lead to inaccurate classification.
[0141] Filtering the portion of the test sequence may include applying one or more rules to filter out low quality reads. For example, if the portion of the test sequence contains more than a threshold number of consecutive identical nucleobases, the portion of the test sequence may be filtered out because it is unlikely that the repeated nucleobases accurately represent a genetic sequence. Continuing with this example, the portion of the test sequence may be filtered out if it contains, for example, more than 10 consecutive instances of cytosine, guanine, adenine, thymine, or combinations of the foregoing. The threshold may vary depending on which nucleobases are being considered. For example, the portion of the test sequence may be filtered out if it contains, for example, more than 10 consecutive instances of thymine, or more than 6 consecutive instances of cytosine. In other examples, other rules may be applied to filter out low quality reads.
[0142] In step 1112, a target is identified in the filtered portion of the test sequence. Step 1112 may be performed in parallel with steps 1108 and / or 1110. That is, additional portions of the test sequence may be received in step 1108 while a portion of the test sequence may be filtered in step 1110 while further targets are identified in the filtered portion of the test sequence in step 1112. Targets may be identified as described above in processes 400 and 700. For example, a determination may be made as to whether any of the target signature fragments obtained in step 1106 are present in the filtered portion of the test sequence. Multiple targets may be identified in the filtered portion of the test sequence. For example, both the SARS-CoV-2 virus and its genetic variants may be identified in the filtered portion of the test sequence. An indication of the number of times each target has been identified in a portion of the test sequence may be stored, for example, in a storage device and / or memory accessible to the processor.
[0143] At step 1114, a determination is made as to whether the test sequence is finished or whether additional portions remain for analysis. For example, step 1114 may include determining whether the sequencer is still outputting sequences for the sample. If the sequence has not been completely sequenced and thus is not terminated (1114 NO), process 1110 returns to step 1108. Steps 1108-1112 are performed repeatedly (and in some instances simultaneously) until it is determined that the entire sequence has been analyzed at steps 1108-1112 (1114 YES), at which point process 1110 proceeds to step 1116. In some instances, steps 1108-1112 are performed at a rate substantially similar to the rate at which the sequencer outputs the sequence, and in some instances at a rate greater than the rate at which the sequencer outputs the sequence. For example, in an example where the sequencer outputs sequences at approximately 4MB / min or more, steps 1108-112 may be performed on at least 4MB / min of information such that analysis of the sequences is performed in substantially real time or near real time with the output of the sequencer.
[0144] At step 1116, a determination is made regarding the probability that one or more targets are present in the test sequence. As described above with respect to step 1112, each portion of the test sequence may be analyzed to determine which targets may be present in the sample based on that portion of the test sequence. In some examples, all targets identified at step 1112 are determined to be present in the sample by virtue of being identified at step 1112. However, in other embodiments, a target may be determined to be present in the sample only if the target is identified at step 1112 at least a threshold number of times. A single determination that a target is present in a sample at step 1112 may not be conclusive when performed on the entire test sequence (rather than a high quality fragment of the test sequence isolated after sequencing is completed), at least since there may be a high probability that the substantially unfiltered test sequence contains errors. Thus, it may be advantageous to determine that a target is present in a sample only if the target is identified at least a threshold number of times in the sample. Furthermore, in some examples, a binary classification (e.g., target present or absent) may not be determined in some examples. Rather, a non-discrete probability that a given target is present may be determined.
[0145] In one example, a target may be determined to be present only if it is identified in step 1112 at least a threshold number of times. For example, the threshold may be 100 instances of the target identified in the sample in step 1112. The threshold may vary based on the target. Multiple thresholds may be implemented for each target. For example, a first threshold may correspond to a low probability that the target is present in the sample, a second threshold may correspond to a medium probability that the target is present in the sample, and a third threshold may correspond to a high probability that the target is present in the sample. Any number of thresholds may be implemented and may vary by target. In some examples, a determination may be made regarding the probability that the target is present in the sample, and a greater number of identifications of the target in step 1112 may generally correspond to a greater probability that the target is present in the sample.
[0146] At step 1118, a probability of presence of each target identified at step 1104 is output. The probability for each target may be expressed, for example, as a binary prediction (e.g., present or absent), a non-discrete probability (e.g., a 98% chance that the target is present), a multi-level prediction (e.g., a prediction that the target is not present, or that the target is low, medium, or high chance of being present), etc. The probability may be expressed differently for different targets, which may vary based on the confidence level that the target is present. For example, if the chance that the target is present is less than 5%, the output at step 1118 may simply indicate that the target is not present. If the chance that the target is present is greater than 99%, the output at step 1118 may simply indicate that the target is present. If the chance that the target is present is between 5 and 99%, the output at step 1118 may indicate a percentage chance that the target is present. In other examples, other formats of output are contemplated. Step 1118 may include outputting the probability(s) to a user, for example, to a user interface accessible to the user. Step 1118 may alternatively or additionally include storing the information in one or more remote or local databases. In other examples, step 1118 may include outputting the information in additional or different manners.
[0147] At step 1120, process 1100 ends. In various examples described above, the nucleic acid sequence may be parsed and used for the target signature fragment, etc. In some examples, the nucleic acid sequence may be converted to an amino acid sequence before performing the parsing, filtering, etc. For example, in process 1000, the target signature fragment may be an amino acid signature, and the remainder of process 1000 may be performed with respect to the amino acid sequence. Thus, references to sequences in the preceding description may refer to nucleic acid sequences, amino acid sequences, a combination of both, etc. Translation of the nucleic acid sequence to an amino acid sequence may be performed during, before, or after the execution of process 1000. Similarly, in process 1100, the test sequence received in step 1108 may be converted to an amino acid sequence before performing the remainder of process 1100.
[0148] 8 is a block diagram of a distributed computer system 800 in which various aspects and functions described above may be implemented. The distributed computer system 800 may include or be coupled to some or all of the components of the systems 100, 500. The distributed computer system 800 may include one or more computer systems. For example, as shown, the distributed computer system 800 includes three computer systems 802, 804, and 806. As shown, the computer systems 802, 804, and 806 are interconnected by a communication network 808 and may exchange data through the communication network 808. The network 808 may include any communication network through which the computer systems may exchange data. To exchange data over the network 808, the computer systems 802, 804, 806 and the network 808 may use a variety of methods, protocols and standards including Token Ring, Ethernet, Wireless Ethernet, Bluetooth, radio signaling, infrared signaling, TCP / IP, UDP, HTTP, FTP, SNMP, SMS, MMS, SS7, JSON, XML, REST, SOAP, CORBA IIOP, RMI, DCOM, and Web Services, among others.
[0149] According to some embodiments, the functions and operations discussed for generating the three-dimensional synthetic perspective may be performed individually and / or in combination on computer systems 802, 804, and 806. For example, computer systems 802, 804, and 806 may support participation in, for example, a collaborative network. In one alternative, a single computer system (e.g., 802) may generate the three-dimensional synthetic perspective. Computer systems 802, 804, and 806 may include personal computing devices such as mobile phones, smartphones, tablets, "phablets," and the like, and may include desktop computers, laptop computers, and the like.
[0150] Various aspects and functions according to the embodiments discussed herein may be implemented as specialized hardware or software executing in one or more computer systems, including computer system 802 shown in FIG. 4. In some embodiments, computer system 802 is a personal computing device specially configured to perform the processes and / or operations described above. As shown, computer system 802 includes at least one processor 810 (e.g., a single-core or multi-core processor), memory 812, bus 814, input / output interfaces (e.g., 816), and storage device 818. Processor 810 may include one or more microprocessors or other types of controllers and may execute a sequence of instructions that manipulate data. As shown, processor 810 is connected to other system components, including memory 812, by interconnect elements (e.g., bus 814). In some examples, processor 810 may be, include, or be coupled to either or both of processors 130, 530. For example, the processor 810, alone or in combination with other devices, may perform any of the processes 300, 400, 600, 700, 1000, and / or 1100.
[0151] Memory 812 and / or storage 818 may be used to store programs and data during operation of computer system 802. For example, memory 812 may be a relatively high performance volatile random access memory such as dynamic random access memory (DRAM) or static memory (SRAM). Additionally, memory 812 may include any device for storing data, such as a disk drive or other non-volatile storage, such as flash memory, solid state, or phase change memory (PCM). In further embodiments, the functions and operations discussed with respect to generating and / or rendering a synthetic three-dimensional view may be embodied in an application executed on computer system 802 from memory 812 and / or storage 818. For example, the application may be made available through an “app store” for download and / or purchase. Once installed or made available for execution, computer system 802 may be specially configured to perform functions related to generating a synthetic three-dimensional view.
[0152] The computer system 802 also includes one or more interfaces 816, such as an input device (e.g., a camera for capturing images), an output device, and a combination input / output device. The interface 816 can receive input, provide output, or both. The storage device 818 can include a computer-readable and computer-writeable non-volatile storage medium on which are stored instructions defining a program to be executed by the processor. The storage device 818 can also include information recorded on or in the medium, which information can be processed by an application. Media that can be used with various embodiments can include, for example, optical disks, magnetic disks or flash memory, SSDs, and the like. Additionally, the aspects and embodiments are not directed to a particular memory or storage system.
[0153] In some embodiments, computer system 802 may include an operating system that manages at least a portion of the hardware components included in computer system 802 (e.g., input / output devices, touch screen, camera, etc.). One or more processors or controllers, such as processor 810, may execute an operating system. The operating system may be, among others, a Windows-based operating system available from Microsoft Corporation (e.g., Windows NT, ME, XP, Vista, 7, 8, or RT), an operating system available from Apple Computer (e.g., MAC OS, including System X), one of many Linux-based operating system distributions (e.g., the Enterprise Linux operating system available from Red Hat, Inc.), the Solaris operating system available from Oracle Corporation, or a UNIX operating system available from a variety of sources. Many other operating systems may be used, including operating systems designed for personal computing devices (e.g., iOS, Android, etc.), and embodiments are not limited to any particular operating system.
[0154] The processor and operating system together define a computing platform upon which applications (e.g., "apps" available from an "app store") may be executed. In addition, various functions for generating and manipulating images may be implemented in a non-programmed environment (e.g., rendering aspects of a graphical user interface or performing other functions when a document created in HTML, XML, or other format is viewed within a browser program window). Furthermore, various embodiments in accordance with aspects of the disclosure may be implemented as programmed or non-programmed components, or any combination thereof. Various embodiments may be implemented, in part, as MATLAB® functions, scripts, and / or batch jobs. As such, the disclosure is not limited to a particular programming language, and any suitable programming language may be used.
[0155] By way of example, computer system 802 is illustrated as one type of computer system on which various functions for generating a three-dimensional synthetic view may be implemented, but aspects and embodiments are not limited to being implemented on the computer system illustrated in Figure 8. Various aspects and functions may be implemented on one or more computers or similar devices having a different architecture or components than that illustrated in Figure 8.
[0156] The examples provided herein allow for the analysis of genetic sequences. In some examples given above, the genetic sequences may be derived from an organism. However, it should be understood that the principles described above are not limited to genetic sequences derived from an organism, but may also be implemented in connection with other sequences, such as synthetic sequences.
[0157] Although several aspects of at least one embodiment have been described above, it will be understood that various changes, modifications, and improvements will readily occur to those skilled in the art. Such changes, modifications, and improvements are intended to be part of this disclosure and are intended to be within the scope of this disclosure. Thus, the foregoing description and drawings are merely examples, and the scope of the disclosure should be determined from proper construction of the appended claims and their equivalents.
Claims
1. A method, implemented by a processor, for analyzing the output of a sequencer in real time, the method comprising: obtaining a plurality of target signature fragments from at least one database, each target signature fragment being derived from a genetic sequence of a genetic target included in a group of designated genetic targets and being a fragment of a genetic sequence registered in said at least one database as distinguishing targets from non-targets; receiving a plurality of portions of a test sequence output by a sequencer that sequences the sample in real time; determining, in real time as the sequencer sequences the sample, whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of the test sequence among the plurality of portions of the test sequence; for each genetic target of the group of genetic targets, determining a respective probability that that genetic target is present in the sample based at least on said determining whether at least one target signature fragment of said plurality of target signature fragments is present in at least one portion of said test sequence among said plurality of portions of said test sequence; and outputting an analysis of the sample, the analysis indicating a respective probability that each genetic target is present in the sample. method.
2. 2. The method of claim 1, wherein the multiple portions of the test sequence include a first portion and a second portion, and the first portion of the test sequence output by the sequencer is received before the second portion of the test sequence is generated by the sequencer.
3. 2. The method of claim 1, further comprising filtering the plurality of portions of the test sequence prior to determining whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of the test sequence among the plurality of portions of the test sequence.
4. The method of claim 3 , wherein filtering the plurality of portions of the test sequence comprises removing low quality portions of the test sequence from the plurality of portions of the test sequence.
5. 5. The method of claim 4, wherein the low quality portion of the test sequence comprises a portion of the test sequence having more than a threshold number of consecutively repeated nucleobases.
6. 2. The method of claim 1, further comprising determining, for each genetic target of the group of genetic targets, a count value indicating the number of times that at least one respective target signature fragment of a respective plurality of target signature fragments corresponding to the respective genetic target is determined to be present in the test sequence.
7. 7. The method of claim 6, wherein for each genetic target of the group of genetic targets, determining a respective probability that the respective genetic target is present in the sample is based on a respective count value of the respective genetic target.
8. 8. The method of claim 7, wherein the respective probability that each genetic target is present in the sample increases as the respective count value of each genetic target increases.
9. determining whether each count value for each genetic target exceeds a respective threshold; and determining that each genetic target is present in the sample based on each count value exceeding the threshold. The method of claim 7.
10. 1. A system for analysing an output of a sequencer in real time, the system comprising: Memory; at least one database configured to store target signature fragments; and at least one processor coupled to the memory and the at least one database, the processor comprising: obtaining a plurality of target signature fragments from the at least one database, each target signature fragment being derived from a genetic sequence of a genetic target included in a group of designated genetic targets and being a fragment of a genetic sequence registered in the at least one database as distinguishing targets from non-targets; receiving a plurality of portions of a test sequence output by a sequencer that sequences the sample in real time; determining, in real time as the sequencer sequences the sample, whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of the test sequence of the plurality of portions of the test sequence; determining, for each genetic target of the group of genetic targets, a respective probability that the genetic target is present in the sample based at least on said determining whether at least one target signature fragment of the plurality of target signature fragments is present in the at least one portion of the test sequence of the plurality of portions of the test sequence; and outputting an analysis of the sample, the analysis indicating a respective probability that each genetic target is present in the sample. system.
11. 11. The system of claim 10, wherein the multiple portions of the test sequence include a first portion and a second portion, and the first portion of the test sequence output by the sequencer is received before the second portion of the test sequence is generated by the sequencer.
12. The system of claim 11 further comprising the sequencer.
13. 11. The system of claim 10, wherein the at least one processor is further configured to filter the plurality of portions of the test sequence before determining whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of the test sequence among the plurality of portions of the test sequence.
14. The system of claim 13 , wherein filtering the plurality of portions of the test sequence comprises removing low quality portions of the test sequence from the plurality of portions of the test sequence.
15. 15. The system of claim 14, wherein the low quality portion of the test sequence comprises a portion of the test sequence having more than a threshold number of consecutively repeated nucleobases.
16. 11. The system of claim 10, wherein the at least one processor is further configured to determine, for each genetic target of the group of genetic targets, a count value indicating the number of times that at least one respective target signature fragment of a respective plurality of target signature fragments corresponding to the respective genetic target is determined to be present in the test sequence.
17. 17. The system of claim 16, wherein for each genetic target of the group of genetic targets, determining a respective probability that the respective genetic target is present in the sample is based on a respective count value of the respective genetic target.
18. 20. The system of claim 17, wherein the respective probability that each genetic target is present in the sample increases as the respective count value of each genetic target increases.
19. The at least one processor further comprises: determining whether each count value for each genetic target exceeds a respective threshold value; and determining that each genetic target is present in the sample based on each count value exceeding the threshold value.
20. The system of claim 17.
20. 1. A non-transitory computer readable medium storing a sequence of computer executable instructions for analyzing in real time an output of a sequencer, the sequence of computer executable instructions including: obtaining a plurality of target signature fragments from at least one database, each target signature fragment being derived from a genetic sequence of a respective genetic target of a group of designated genetic targets and being a fragment of a genetic sequence registered in said at least one database as distinguishing targets from non-targets; receiving a plurality of portions of a test sequence output by a sequencer that sequences the sample in real time; determining, in real time as the sequencer sequences the sample, whether at least one target signature fragment of the plurality of target signature fragments is present in at least one portion of the test sequence of the plurality of portions of the test sequence; determining, for each genetic target of the group of genetic targets, a respective probability that the genetic target is present in the sample based at least on said determining whether at least one target signature fragment of the plurality of target signature fragments is present in the at least one portion of the test sequence of the plurality of portions of the test sequence; and outputting an analysis of the sample, the analysis indicating a respective probability that each genetic target is present in the sample. Medium.
Citation Information
Patent Citations
System, method and computer readable medium for rapid DNA identification
US20160132640A1
Method for identification of nucleic acid signatures
US20180089365A1
