Method for functionally screening a biological sequence fragment

Abstract

The present invention relates, in part, to methods for accurately and reliably detecting biological sequences corresponding to particular functions.

Description

[0001] Related applications

[0002] This application claims the benefit of U.S. Provisional Application Serial No. 62 / 965,138, filed January 23, 2020, pursuant to 35 USC §119(e), the disclosure of which is incorporated herein by reference in its entirety. Technical Field

[0003] This invention relates in part to a method for detecting biological sequences corresponding to specific biological functions while minimizing false detections of sequences with irrelevant functions. Background Technology

[0004] The government strongly recommends that the industry screen nucleic acid synthesis orders to prevent the construction of biological weapons [Diggans, J. and E. Leproust. Frontiers in Bioengineering and Biotechnology 7 (April): 86 (2019)]. Many companies voluntarily screen reagents identified by national and international organizations [see International Gene Synthesis Consortium (2017) "Harmonized Screening Protocol V2 / / genesynthesisconsortium.org / wp-content / uploads / IGSCHarmonizedProtocoll l-21-17.pdf."]. Current screening methods rely on similarity search algorithms [Altschul et al. Journal of Molecular Biology 215(3):403-10(1990)] to identify sequences similar to those from known biological weapons. These algorithms cannot screen out small fragments of nucleic acids that can be assembled into larger fragments. Many compliant (innocent) sequences are similar enough to be identified as dangerous by similar searches, which produces false positives that require human intervention and excludes automated screening. Automated screening methods applicable to benchtop nucleic acid synthesizers and assemblers are not available, and these methods do not necessarily rely on human experts to handle false positives. Summary of the Invention

[0005] According to one aspect of the invention, a method is provided for evaluating biological sequences capable of performing a pre-selected function, the method comprising: (a) pre-selecting a biomolecule capable of performing the desired function; (b) preparing a test sequence database comprising a plurality of sequence fragments of the pre-selected biomolecule, wherein the pre-selected sequence fragments have a predetermined length; (c) fragmenting one or more test biomolecules in the test sequence database to a length equal to the predetermined length of the sequence fragments of the pre-selected biomolecules; (d) detecting the presence or absence of a sequence match between the sequence of at least one fragment of the fragmented test biomolecule and the sequence of at least one of the plurality of sequence fragments of the pre-selected biological sequence; and (e) acting in response to the detection in (d), wherein the detection in (d) provides for the evaluation of the test biomolecule. In some embodiments, acting in response to (d) includes one or more of the following: preventing the synthesis of the test biomolecule, allowing the synthesis of the test biomolecule, sequencing one or more polynucleotide molecules, DNA sequencing, DNA molecule design, polypeptide sequence determination, and other sequence identification steps. In some embodiments, the method further includes identifying one or more sequence fragments of a preselected biological sequence in a test sequence database, said one or more sequence fragments matching, respectively, one or more sequence fragments of a second biomolecule having a biological function unrelated to the target biological function of the preselected biomolecule, and removing the identified sequence fragments from the test sequence database. In some embodiments, if the presence of a sequence match is detected in (d), this action includes preventing the synthesis of the test biomolecule. In some embodiments, the means of preparing a test sequence database include: (a) screening multiple sequence fragments of preselected biological sequence molecules against at least one control sequence database, wherein the control sequence database contains multiple control sequence fragments of at least one molecule capable of performing a purpose unrelated to the purpose function of the preselected biomolecule; (b) identifying a match between sequence fragments of the multiple sequence fragments of the preselected biomolecule and sequence fragments in the control sequence database, the sequence fragments in the control sequence database being fragments of biomolecules identified as capable of performing a purpose unrelated to the purpose function of the preselected biomolecule; and (c) removing such preselected biological sequence fragments from the test sequence database that are identified as matching such biological sequence fragments, the biological sequences being identified as capable of performing a purpose unrelated to the purpose function of the molecule (which is capable of performing the purpose function). In some embodiments, the preselected biomolecule is a polynucleotide. In some embodiments, the sequence of the biomolecule is a full-length nucleic acid sequence of a polynucleotide or a portion thereof. In some embodiments, the full-length nucleic acid sequence encodes a protein.In some embodiments, the predetermined length of the sequence fragment of the preselected polynucleotide molecule is 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200 or more nucleotides. In some embodiments, the preselected biomolecule includes a polypeptide. In some embodiments, the amino acid sequence of the preselected biomolecule is the full-length amino acid sequence of the polypeptide, or a portion thereof. In some embodiments, the predetermined length of the sequence fragment of the preselected polypeptide molecule is 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 2, 93, 94, 95 or more amino acids. In some embodiments, the multiple sequence fragments of the preselected biomolecule: (1) contain all or most of a possible fragment of a biomolecule capable of achieving the intended function or at least one key fragment of said biomolecule, and (2) do not contain sequences present in biomolecules capable of achieving functions unrelated to the function of the preselected biomolecule. In some embodiments, the control sequence database contains multiple control sequence fragments of at least one molecule capable of performing a target function unrelated to the target function of the preselected biomolecule. In some embodiments, the control sequence database contains multiple control sequence fragments of at least one molecule that cannot perform the target function of the preselected biomolecule. In some embodiments, the predetermined length is the same for all fragments of the preselected biomolecule. In some embodiments, the predetermined length of the sequence fragments of the preselected biomolecule includes more than one length. In some embodiments, the test sequence database contains one or more sequence fragments randomly or pseudo-randomly selected from molecular sequences known to perform functions different from the target function of the preselected molecule. In some embodiments, the randomly or pseudo-randomly selected sequence fragments are biased towards sequence regions with higher homology to functionally or phylogenetically related sequences. In some embodiments, the test sequence database also contains sequences that are functional equivalents of multiple sequence fragments of the preselected biomolecule. In some embodiments, the means for identifying functional equivalents include computational means. In some embodiments, the means for identifying functional equivalents include experimental means.In some embodiments, the computational means for selecting functional equivalents contained in the test sequence database includes using a classifier based on experimental data to evaluate the accuracy of the computational means. In some embodiments, the means for selecting functional equivalents contained in the test sequence database includes including the inclusion of a minimum number of sequences calculated to achieve a predetermined likelihood of successfully preventing the test sequence from escaping detection. In some embodiments, the means for selecting functional equivalents contained in the test sequence database includes random selection methods or pseudo-random selection methods. In some embodiments, the identity of all sequence fragments of the test sequence database and one or both of the test biomolecules is protected. In some embodiments, the means of protection include applying a cryptographic hash function, wherein the cryptographic hash function deterministically maps sequence data to a fixed-size bit string using a one-way function. In some embodiments, the application of the cryptographic hash function cannot be reversed without a brute-force search of all possible sequences input to the test sequence database. In some embodiments, the application of the cryptographic hash function also includes using one or more information keys, which must be accessed to attempt a brute-force search. In some implementations, the application of the cryptographic hash function requires keys from multiple independent sources that must collaborate to compute the hash without any server gaining access to the sequence data. In some implementations, the independent sources include independent computer servers. In some implementations, the method further includes dividing the prepared test sequence database into two or more partial test sequence databases, with the test sequence database prepared for detecting the presence or absence of sequence matches being one of two or more partial test sequence databases. In some implementations, if a sequence match is detected between a partial test sequence database and one or more fragments of the test biomolecule, the method further includes using another of the two or more partial test sequence databases to detect the presence or absence of the sequence. In some implementations, the test sequence database comprises a portion of a larger sequence fragment database, such that the fragments contained in the test sequence database can be frequently rotated or rotated upon the discovery of a match.

[0006] According to another aspect of the invention, a method is provided for identifying biological sequences capable of performing a pre-selected function, the method comprising: (a) pre-selecting biomolecules, wherein the pre-selected biomolecules are capable of performing the intended function; (b) preparing a test sequence database comprising a plurality of sequence fragments of the pre-selected biomolecules, wherein the pre-selected sequence fragments have a predetermined length; (c) fragmenting one or more test biomolecules in the test sequence database to a length equal to the predetermined length of the sequence fragments of the pre-selected biomolecules; and (d) detecting the presence or absence of a sequence match between at least one fragment of the fragmented test biomolecule and at least one of the plurality of sequence fragments of the pre-selected biological sequence; wherein the means of preparing the test sequence database includes: (i) screening a pre-selected biomolecule against at least one control sequence database. The method involves: (i) selecting multiple sequence fragments of a preselected biomolecule, wherein a control sequence database contains multiple control sequence fragments of at least one molecule capable of performing a target function unrelated to the target function of the preselected biomolecule; (ii) identifying the presence of a match between a sequence fragment of the multiple sequence fragments of the preselected biomolecule and a sequence fragment in the control sequence database, the sequence fragment in the control sequence database being a fragment of a biomolecule identified as capable of performing a function unrelated to the target function of the preselected biomolecule; and (iii) removing such a sequence fragment of the preselected biomolecule identified as matching a sequence fragment of such a biomolecule identified as capable of performing a target function unrelated to the target function of the molecule. In some embodiments, the preselected biomolecule is a polynucleotide. In some embodiments, the sequence of the preselected biomolecule is a full-length nucleic acid sequence of a polynucleotide or a portion thereof. In some embodiments, the full-length nucleic acid encodes a protein. In some embodiments, the predetermined length of the sequence fragment of the preselected polynucleotide molecule is 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200 or more nucleotides. In some embodiments, the preselected biomolecule comprises a polypeptide molecule. In some embodiments, the sequence of the preselected biomolecule is the full-length amino acid sequence of the polypeptide, or a portion thereof.In some embodiments, the predetermined length of the sequence fragment of the preselected polynucleotide molecule is 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 2, 93, 94, 95 or more amino acids. In some embodiments, the multiple sequence fragments of the preselected biomolecule: (1) contain all or most of the possible fragments of the biomolecule capable of achieving the intended function, and (2) do not contain sequences not found in biomolecules capable of achieving functions unrelated to the function of the preselected biomolecule. In some embodiments, the control sequence database contains multiple control sequence fragments of at least one molecule capable of performing a target function unrelated to the target function of the pre-selected biomolecule. In some embodiments, the control sequence database contains multiple control sequence fragments of at least one molecule that cannot perform the target function of the pre-selected biomolecule. In some embodiments, a single length is a predetermined length of the sequence fragment of the pre-selected biomolecule. In some embodiments, the sequence fragments of the pre-selected biomolecule have more than one predetermined length. In some embodiments, the test sequence database contains one or more sequence fragments randomly selected from molecular sequences known to perform functions different from the target function of the pre-selected molecule. In some embodiments, the randomly selected sequence fragments are biased towards conserved regions. In some embodiments, the test database also contains sequences that serve as functional equivalents of the multiple sequence fragments of the pre-selected biomolecule. In some embodiments, the identity of the test sequence database and all sequence fragments of one or both of the test biomolecules is protected. In some embodiments, the means of protection include applying a cryptographic hash function, wherein the cryptographic hash function deterministically maps sequence data to a fixed-size bit string using a one-way function. In some implementations, the application of the cryptographic hash function cannot be reversed without an exhaustive search of all possible sequences input into the test sequence database. In some implementations, the application of the cryptographic hash function also includes using one or more information keys, which must be accessed to attempt an exhaustive search. In some implementations, the application of the cryptographic hash function requires keys from multiple independent sources that must cooperate to compute the hash without any server gaining access to the sequence data. In some implementations, the independent sources comprise independent computer servers. In some implementations, the method further includes dividing the prepared test sequence database into two or more partial test sequence databases, and the prepared test sequence database for detecting the presence or absence of sequence matches is one of two or more partial test sequence databases.In some embodiments, if a sequence match is detected between a partial test sequence database and one or more fragments of a test biomolecule, the method further includes using another of two or more partial test sequence databases to detect the presence or absence of the sequence. In some embodiments, the test sequence database comprises a portion of a larger database of sequence fragments, such that the fragments contained in the test sequence database can be frequently rotated or rotated upon the discovery of a match. In some embodiments, the method further includes acting in response to detection, wherein this action includes one or more of the following: preventing the synthesis of the test biomolecule, allowing the synthesis of the test biomolecule, sequencing one or more polynucleotide molecules, DNA sequencing, DNA molecule design, determining the amino acid sequence of a polypeptide, and additional sequence identification steps. In some embodiments, if a sequence match is detected, this action includes preventing the synthesis of the test biomolecule. In some embodiments, the method further includes identifying one or more sequence fragments of a preselected biological sequence in the prepared test sequence database, said one or more sequence fragments respectively matching one or more sequence fragments of a second biomolecule having a biological function unrelated to the intended function of the preselected biomolecule, and removing the identified sequence fragments from the test sequence database.

[0007] In another aspect of the invention, a test sequence database prepared by any embodiment of any of the methods described above is provided.

[0008] In another aspect of the invention, a method for evaluating biological sequences using the aforementioned test sequence database is provided. In some embodiments, the evaluation includes determining whether to permit or prevent the synthesis of the biological sequence being evaluated. Attached Figure Description

[0009] Figure 1 This diagram illustrates how nucleic acid and peptide sequences can be broken down into fragments of a certain length to detect precise matches within a unique database of potential biological weapons fragments. This database may contain known sequence fragments from potential biological weapons and / or calculated functional equivalents, but not fragments that match functionally unrelated sequences from public databases.

[0010] Figure 2 This diagram illustrates how to hash sequences and database contents to allow filtering while avoiding providing any information in cleartext.

[0011] Figures 3A to 3B It provides a description of how to create a sequence database corresponding to potential biological weapons for precise matching and comparison with nucleic acid sequences to be examined. Figure 3AA schematic diagram is provided illustrating how to calculate functionally equivalent fragments in a list sorted by functional probability using fragments of nucleic acid or peptide of appropriate size. Figure 3B The entries from the sorted list are shown and examples are given of how these entries can be included in the database in a deterministic, random, or biased-to-random manner up to the random adversarial threshold.

[0012] Figures 4A to 4B A graphical representation of how to randomly select (optionally in a way that favors conserved regions) fragments from potential biological weapons is provided, which can be used to reliably detect sequences corresponding to functionally similar biological weapons. Figure 4A Five fragments and calculated functionally equivalent variants are depicted, along with descriptions of naive or sophisticated “attacks” that attempt to evade detection by introducing mutations throughout the biological weapon sequence. Figure 4B This example illustrates the failure of attempts to evade filtering because the adversary does not know which fragments are contained in the database or how many functional variants of those fragments are included, and tries to guess the variants contained in the database when trying to avoid the function becoming invalid due to too many mutations.

[0013] Figure 5 It depicts the estimated number of false positives per year for global DNA synthesis levels over time, assuming a database size of approximately 1 billion fragments and 57-base-pair nucleic acid sequences or 19-amino acid peptide sequences.

[0014] Figure 6 This is a schematic flowchart that provides an overview of one embodiment of the invention and illustrates the use of an exemption list.

[0015] Figure 7 The RAT screening plot illustrates how different segments of a gene have different fitness costs when mutated: some can be changed to almost anything, others allow some substitutions but would otherwise disrupt function, while still others exhibit such a gradient, with most individual mutations producing small costs that increase with the addition of more mutations. SecureDNA is a screening database prepared using one embodiment of the method of the present invention.

[0016] Figure 8 A schematic diagram of an implementation is provided for phage-particle-based selection to measure the fitness of sequence variants of M13 (an exemplary virus infecting Escherichia coli). Enrichment / de-enrichment relative to the wild type corresponds to variant fitness.

[0017] Figure 9 An example is provided of a series of experimental steps used in one embodiment of the invention to generate fitness data for sequence variants of Ml3.

[0018] Figure 10 This is a diagram illustrating the effects of repeated selection extrusion and infection on a variant library of the M13 pIII peptide PQSVECRPFVFGAGKPYEF.

[0019] Figure 11 Enrichment plots of all single mutants of alanine at position 13 in the M13 pIII polypeptide PQSVECRPFVFGAGKPYEF (SEQ ID NO:23) are provided. The results indicate that all 19 mutations are permitted at this position.

[0020] Figure 12 An enrichment map of all single mutants at position 1 of the M13 pIII peptide PQSVECRPFVFGAGKPYEF (SEQ ID NO:23) is shown. The results indicate that mutations at this position are not permitted.

[0021] Figure 13 This provides a chart predicting all amino acids in a peptide PQSV_funtrp_analysis using the funtrp procedure. The N column corresponds to the likelihood that the position is neutral and accepts most or all substitutions with minimal loss of function. The R column corresponds to the likelihood that the position is a rheostat, which accepts a certain number of mutations that reduce function to varying degrees. The T column corresponds to the likelihood that the position is a toggle that does not allow mutations without loss of function.

[0022] Figure 14 Such a graph is provided, showing the Receiver Operating Characteristic (ROC) curves of weak classifiers based on the tools FUNTRP and BLOSUM62, as an evaluation of basic biological fact data obtained in the laboratory. The relevant data is a sequence fragment window of the pIII protein from M13, a filamentous virus that infects E. coli (as described in Example 7). The choices used are as follows... Figure 8 As shown in the figure, the enrichment / de-enrichment scores of NGS at point 1 and NGS at point 6 are compared. The curve indicates that, for the range of sequences tested, 90% of true positives can be predicted at the cost of 50% false positives.

[0023] Figure 15 Provides the same classifier (e.g.) Figure 14A graph of another receiver operating characteristic (ROC) curve (in the middle) compares the enrichment / de-enrichment scores of NGS at point 1 and point 4. The ROC curve similarly indicates that repeated selection is not required. Detailed Implementation

[0024] Some aspects of this invention include methods and systems for reliably and efficiently detecting sequences corresponding to preselected biological functions (also referred to herein as "functional sequences") while minimizing the detection of functionally irrelevant sequences (also referred to herein as "irrelevant sequences"). In some embodiments, the method of this invention includes detecting functional sequences of nucleic acids. In some embodiments, the method of this invention includes detecting functional sequences of peptides. The term "peptide" as used herein is used interchangeably with the term "protein." One embodiment of the detection system of this invention may include a database of test sequences as described herein.

[0025] The method of the present invention can be used to detect nucleic acid or peptide sequences corresponding to specific key biological functions, so as to reliably identify sequences encoding the function and minimize the possibility of incorrectly identifying sequences that do not correspond to the function.

[0026] Certain embodiments of the present invention can be used to prevent individuals interested in sequences deemed undesirable (also referred to herein as “adversaries”) from using such sequences in synthesis to avoid detection. Randomly selecting fragments from functional sequences prevents adversaries from knowing which fragments will be screened, forcing them to include mutations throughout the test sequence in an attempt to evade detection. If adversaries do not include enough mutations at a particular fragment, their sequence may match one of the calculated functional variants contained in the test sequence database of the present invention. The more fragments included, and the more calculated functional variants of those fragments, the greater the likelihood of detection. If an adversary includes too many mutations throughout their test sequence, they will no longer perform the desired function [Gray et al. Genetics 207(1):53-61(2017); Jackson et al. PLoS One 12(4):e0164905(2017) and Pokusaeva et al. PLoS Genetics 15(4):e1008079(2019)].

[0027] Test sequence database

[0028] In some embodiments of the invention, the detection method includes searching for and / or identifying sequences that match a sequence database. In some embodiments, the sequence database, also referred to herein as a “test sequence database,” contains multiple sequence fragments of preselected biomolecules. In some embodiments, the preselected biomolecules are selected at least partially because they are capable of achieving the intended function. The preselected biomolecules may be polypeptide molecules or polynucleotide molecules, and may be capable of achieving the intended function. Some non-limiting examples of biomolecules capable of achieving the intended function include sequences corresponding to viruses capable of human-to-human transmission (e.g., but not limited to Ebola virus); and sequences encoding toxins capable of killing mammalian cells at very low doses (e.g., but not limited to ricin). Other biomolecules capable of achieving the intended function are known in the art, and such sequences may be included in some embodiments of the methods of the invention.

[0029] The test sequence database can be prepared in such a manner that it contains multiple sequence fragments of preselected biomolecules; such fragments may also be referred to herein as "preselected sequence fragments". In some embodiments of the invention, the preselected sequence fragments in the test sequence database have a predetermined length. In some embodiments where the preselected biomolecule is a polynucleotide, the predetermined length of the preselected sequence fragment is 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 300 or more nucleotides, comprising all integers from 15 to 300. In some embodiments where the preselected biomolecule is a polypeptide, the predetermined length of the preselected sequence fragment is: 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 2, 93, 94, 95, 100, 110, 120, 130, 140, 150 or more amino acids, comprising all integers from 7 to 150. In some embodiments, the test sequence database contains preselected sequence fragments of the same predetermined length. In some embodiments, the test sequence database contains preselected sequence fragments of different predetermined lengths.

[0030] In some embodiments of the invention, the plurality of sequence fragments of the preselected biomolecule includes all or a significant portion of possible fragments of the biomolecule capable of achieving the intended function. In some embodiments of the method of the invention, the plurality of sequence fragments of the preselected biomolecule does not include sequences found in the biomolecule that can achieve functions unrelated to the function of the preselected biomolecule. The term "plural" as used herein means more than one, for example, it may mean at least 2, 3, 4, 5, 6, 7, 8, 9, 10 or more.

[0031] Figure 1 and 2 Schematic diagrams illustrating certain embodiments of the method of the present invention are provided. Figure 1 Examples illustrate methods for breaking down nucleic acid and peptide sequences into fragments of a certain length to detect precise matches in a database of fragments specific to potential biological weapons. In some embodiments of the invention, the database may contain known sequence fragments from potential biological weapons and / or calculated functional equivalents, but not fragments that match functionally unrelated sequences from public databases. Figure 2 This illustrates how, in certain embodiments of the invention, sequences and database contents to be filtered are hashed to allow filtering while avoiding providing any information in plaintext.

[0032] Some methods for creating test sequence databases

[0033] In some embodiments of the invention, the means of preparing a test sequence database includes screening multiple sequence fragments of preselected biological sequence molecules against at least one control sequence database, wherein the control sequence database contains multiple control sequence fragments of at least one molecule capable of performing a desired function, the function being unrelated to the desired function of the preselected biomolecule. The means of preparing the test sequence database may further include identifying the presence of matches between sequence fragments of the multiple sequence fragments of the preselected biomolecule and sequence fragments in the control sequence database, the control sequence database being fragments of biomolecules identified as capable of performing a function unrelated to the desired function of the preselected biomolecule. As used herein, the term "control sequence database" means a database containing one or more sets of sequences unrelated to the sequence sought by the detection system, the inclusion of which in the test sequence database would result in false positive matches. Some non-limiting examples include sequences of all plasmids in GenBank, the European Nucleotide Archive, and the Addgene repository requested by at least 25 laboratories.

[0034] Furthermore, the means of preparing a test sequence database may also include removing one or more sequence fragments of such preselected biological sequences from the test sequence database, which are identified as matching sequence fragments of such biological sequences, said biological sequences being identified as capable of achieving a purpose function unrelated to the purpose function of the molecule capable of achieving the purpose function. As used herein, the term "biological sequence" refers to a molecule found in a biological system, and some non-limiting examples of biological sequences are DNA sequences, RNA sequences, gene sequences, polynucleotide sequences, protein sequences, polypeptide sequences, amino acid sequences, and nucleic acid sequences.

[0035] In some embodiments of the method and system of the present invention, the test sequence database contains fragments of randomly selected functional sequences. In some cases, the ordination list of sequences that are predicted to be functionally equivalent to those contained in the test sequence database is computed using methods known in the art (see, for example: Bromberg, U, & B. Rost Nucleic Acids Res. 35, 3823-3835 (2017); Miller et al. Sci. Rep. 7, 41329 (2017); Miller, M. et al. Nucleic Acids Res. 47, e142 (2019); Choi, Y. et al. PLoS One 7 (10), e46688 (2012); Hopf, TA et al. Nat. Biotechnol. 35, 128-135 (2017); Gray, VE et al. Cell Syst. 6, 116-124, e3 (2018); and Riesselman, AJ et al. Nat. Methods [15,816-822 (2018), the contents of which are incorporated herein by reference in their entirety]. Some non-limiting examples of a minimal number are: for testing purposes, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 or more of the list of functionally equivalent sequences may be included in the test sequence database. For example, such functionally equivalent sequences may be included as control sequences.

[0036] In some embodiments of the method of the present invention, one or more calculated equivalent functional sequences may be randomly or in a biased random manner selected from a sorted list of predicted functionally equivalent sequences calculated using methods known in the art and included in the test sequence database of the present invention. In some embodiments, pre-screened randomly selected segments and calculated equivalent functional segments are used to match known irrelevant sequences present in a database, a non-limiting example of which is GenBank, to ensure that the test sequence database does not contain segments that erroneously suggest irrelevant sequences.

[0037] Some embodiments of the present invention include a pre-screening step in which sequences unrelated to the sequence of a pre-selected biomolecule are tested. In some embodiments of the present invention, the rate of false identification of irrelevant sequences from a set of known sequences included in the pre-screening step is 0% before the test sequence database is populated. The rate of random false identification of unknown or irrelevant sequences not included in the pre-screening step varies with the fragment length and number of fragments contained in the database, wherein the false identification rate for each fragment corresponds to the total number of nucleic acid or peptide sequences of a defined length for each fragment. Using the apparatus, system, and method of the present invention, true functional sequences can be reliably identified at rates of 90%, 95%, 99%, 99.9%, or 100%, including all percentages within the provided range, wherein the exact rate depends on the number of randomly selected fragments and equivalent functional sequence fragments contained in the test sequence database.

[0038] Figure 3A Segment B illustrates a non-limiting example of preparing the database of the present invention. In this example, a database corresponding to sequences of potential biological weapons was prepared for precise matching and comparison with nucleic acid sequences to be examined. Figure 3A A schematic diagram is provided showing how functionally equivalent fragments are calculated from a list sorted by functional probability using fragments of nucleic acids or peptides of a predetermined size. Figure 3B Some items from such a sorted list are shown, and examples are given of those items that can be included in the database in a deterministic, random, pseudo-random, or biased-to-random manner, up to the randomness adversarial threshold.

[0039] In another example of one embodiment of the present invention, Figure 4A Figures B to B illustrate a graphical representation of how fragments can be randomly selected from potential biological weapons in a manner biased towards conserved regions, which can be used to reliably detect sequences corresponding to functionally similar biological weapons. As used herein, the term "biased towards conserved regions" refers to the selection of sequences exhibiting a higher level of homology with sequences of related genes and organisms. Such homology is generally associated with greater functional importance. Figure 4A Five fragments and calculated functionally equivalent variants are shown, along with descriptions of initial or sophisticated “attacks” that attempt to evade detection by introducing mutations throughout the biological weapon sequence. Figure 4BExamples illustrate the failure of attempts to evade screening, as adversaries are unaware of which fragments are contained in the database or how many functional equivalents (also referred to as functional variants here) are present, and attempt to avoid presentations that would render the function ineffective due to too many mutations, speculating that the database contains variants. The term "functional equivalent" as used herein with respect to the first nucleotide or polypeptide sequence refers to a second nucleotide or polypeptide sequence that can replace the first nucleotide or polypeptide sequence without significantly compromising the function of the overall sequence, the biomolecule, or the molecule encoded by that sequence.

[0040] A database prepared using one embodiment of the method of the present invention is intended to allow the identification of potentially harmful nucleotide and / or polypeptide sequences and provide an opportunity to prevent the generation of potentially harmful nucleotide and / or polypeptide sequences. When screening sequences using a database prepared using one embodiment of the present invention, the possibility of false positive results is very low. For example, Figure 5 It shows an estimated level of global DNA synthesis over time, the projected number of false positives per year, assuming a database size of approximately 1 billion fragments and 57-base-pair nucleic acid sequences or 19-amino acid peptide sequences.

[0041] Exemption List

[0042] As used herein, the terms “exempt sequence” and “exempt list” refer to sequences for which an individual and / or laboratory is expressly authorized to use. Therefore, when requesting the synthesis of a sequence, the requesting individual or laboratory may provide the synthesis facility with a list of sequences that the individual and / or laboratory is permitted to possess and / or use. As a non-restrictive example, a laboratory may be permitted to use sequence “X” for work that is considered a hazardous sequence but is necessary for the laboratory’s research to develop treatments or vaccines against organisms containing sequence “X”. Other individuals and / or laboratories are not permitted to synthesize or use sequence “X”, but it will be considered an exempt sequence for the permitted laboratory and will be on the laboratory’s exemption list.

[0043] Figure 6 An overview of one embodiment of the invention is provided, along with a flowchart illustrating how an exemption list works. For example, laboratories typically require permission from their institution's biosafety committee or other authoritative body to use certain reagents. Some embodiments of the fully automated screening system prepared using the methods of the present invention will reveal that such laboratories are permitted to obtain DNA corresponding to the genes and genomes allowed for their use without any human intervention. In a non-limiting example, the genes and genomes listed in the biosafety committee authorization report all have relevant... IDs are available for all genes and genomes. The exemption list contains all GenBank IDs of genes and genomes for which the laboratory is explicitly authorized to use. This information can be used to identify genes and genomes permitted for use with respect to the screening system of this invention. In some embodiments of the system of this invention, each laboratory wishing to synthesize sequences needs to send its exemption list to the screening system of this invention, which hashes each GenBank ID once using a distributed oblivious multiparty server system, and then hashes it again using the laboratory's unique ID as salt. This ensures that all laboratories have different and unique hashes, which maintains the privacy of the list and prevents adversaries with copies of the database and knowledge of which laboratories are working on which genes and genomes from determining which hash corresponds to which gene or genome. Thus, adversaries can be prevented from using this information to determine what is present in the database. When a harmful gene or genome is entered into the database for protection by the operator of the screening system of this invention, the corresponding GenBank ID is hashed once using the distributed oblivious multiparty server system, and the corresponding GenBank ID is associated with each hashed sequence fragment of that gene or genome in the database. If a user places an order and the system of this invention detects that a DNA fragment exists in a harmful database, the database uses the customer's lab ID to match the relevant (hash-based) DNA fragment. The ID is hashed, and then checked to see if a hash matching an entry in the customer exemption list (a similar hash) is generated. If so, the order is processed and the requested sequence is synthesized. If not, the system of the present invention rejects the order and the event can be logged.

[0044] Suitability Cost

[0045] In some embodiments of the method of the present invention, it includes assessing how fragments of a gene or gene product sequence have different fitness costs when mutated. Figure 7The RAT screening diagram illustrates how some segments of a gene can be altered to almost anything, while other segments will tolerate minor substitutions, but these substitutions will also disrupt the function of the gene or gene product. Other segments of the sequence still exhibit a gradient where most individual mutations produce small costs, with the cost increasing with the addition of more mutations. The stochastic adversarial thresholding method described herein for the method of the invention works by using a classifier algorithm to predict the function of mutants for each segment and including numerous wild-type sequences and predicted mutant sequences from a database. To obtain the functional form of a harmful gene or genome protected by the system of the invention, the adversary must select and issue synthetic orders with wild-type sequences or variants for each window of the gene or genome. For an order to pass, all these guesses must not exist in the database. Because the adversary does not know which windows are protected, they must guess the variants for each window, which introduces potential fitness costs and the risk of being detected in each window. For the resulting sequence to be functional, the total cost of the system must not exceed a minimum threshold. The higher the proportion of variants protected within a given window and the more windows protected, the lower the likelihood that the order will escape detection. Because this method can be fully automated, it can be applied to next-generation desktop DNA synthesizers as well as central suppliers.

[0046] Some applications of the method

[0047] Certain embodiments of the present invention can be used for biosafety applications. For example, while not intended to limit the methods of the present invention, they can be used to detect functional sequences corresponding to bioweapons in DNA synthesis orders, thereby preventing such synthesis and rejecting those orders. Another non-limiting embodiment of the method of the present invention includes detecting functional sequences corresponding to bioweapons in DNA sequencing results. Another non-limiting embodiment of the present invention includes detecting functional sequences corresponding to bioweapons from a set of sequences input into a DNA design and analysis software program.

[0048] Certain aspects of this invention allow for the efficient calculation of whether a sequence is effective. Time corresponding to O(log(N)) is considered the gold standard for the best fast algorithms, where, in the context of this invention, N corresponds to the number of segments in the database (Cormen, TH, et al., 2009. Introduction to Algorithms. MIT Press.). Some data structures allow for exact match lookups in time corresponding to O(1); because this invention relies on exact match lookups, certain embodiments of this invention allow for similar efficiency.

[0049] Some embodiments of the present invention allow for the automated screening of functional sequences without human intervention. For example, in some embodiments of the method according to the invention, nucleic acid or peptide synthesis machines can be programmed to automatically screen and reject sequence synthesis orders containing functional sequences derived from lists of prohibited organism and toxin genes, such as those from the U.S. Federal Select Agent Program (FSAP) and the Australia Group treaty for harmonized export control.

[0050] Molecules and sequences used for screening

[0051] In some embodiments of the invention, test sequences are screened against a test sequence database in the method or system of the invention. The term "screened against" means "compared to". In a non-limiting example, the target sequence to be synthesized is a test sequence, and it is screened using the method and / or system of the invention. Screening against the test sequence database of the invention provides information that can help determine the action to be taken regarding the test sequence, such as, but not limited to, whether to allow or block the sequence to be synthesized. Other actions that can be informed by applying an embodiment of the method of the invention to the results of the test sequence include, but are not limited to, sequencing one or more polynucleotide molecules, DNA sequencing, DNA molecule design, and additional sequence identification steps. Various means of evaluating DNA molecule design, DNA and / or protein sequence sequencing are known in the art and can be applied as part of actions taken based at least in part on information generated by an embodiment using the test database of the invention.

[0052] In some embodiments of the invention, the test biomolecule is fragmented into one or more of the following: multiple, some, and all of fragments that may overlap by a shift of a base pair or an amino acid of a desired length, used for comparison with fragments of equivalent size from a related sequence. The length of the fragmented sequence of one or more test biomolecules is equal to a predetermined length of sequence fragments of preselected biomolecules in a test sequence database. The test biomolecule is a molecule evaluated / tested using the test sequence database of the invention. For example, although not intended to be limiting, the test biomolecule can be a polynucleotide that an individual or laboratory wants to synthesize or that has been produced by a service provider or synthesizer.

[0053] Sequence identity protection

[0054] In some methods and systems of the present invention, the identity of sequence fragments of one or both of the test sequence database and the test biomolecule is protected. The term “protected” as used herein means that a user of the method of the system of the present invention is prevented from identifying the sequence fragment or the test biomolecule. For example, methods known to those skilled in the art can be used to “hash” the sequence fragments to be screened to produce a one-to-one mapping of information that cannot be easily reversed. Including equivalent hash fragments from related sequences in the test sequence database allows for reliable database lookups and detections without disclosing the identity of the sequences. A variety of means known in the art can be used to protect the identity of sequences. A non-limiting example of a protection means includes the application of cryptographic hash functions, which use a single-factor function to deterministically map sequence data to a fixed-size bit string (see, for example: Cormen, T. FL, et al., 2009. Introduction to Algorithms. MIT Press.). Cryptographic hash functions are used in the art, and methods known in the art can be used to include cryptographic hash functions in the methods and systems of the present invention. In some embodiments of the invention, a cryptographic hash function is selected and applied that cannot be reversed or deciphered without an exhaustive search of all possible sequences input into the test sequence database. In some embodiments of the invention, the applied cryptographic function also includes the use of one or more information keys, which must be accessed to attempt an exhaustive search. Including one or more such information keys restricts the user's ability to access the test sequence database and test the identity of sequence fragments of one or both of the biomolecules. It should be understood that other means for protecting the identity of sequence fragments may also be used in conjunction with embodiments of the method of the invention. See, for example: Yao, AC27th Annual Symposium on Foundations of Computer Science (SFCS 1986), Toronto, ON, Canada, 1986, pp. 162–167; and I. Damgard, I. '89 Proceedings, Lecture Notes in Computer Science, Vol. 435, edited by G. Brassard, Springer-Verlag, 1990, pp. 416–427, the contents of which are incorporated herein by reference in their entirety.

[0055] Example

[0056] Example 1

[0057] A database of test sequences was constructed by selecting all possible fragments from lists of prohibited organisms and toxin genes, such as those from the U.S. Federal Selective Pharmacy Program (FSAP) and the Australian Group Treaty. This database was pre-screened against known databases (such as GenBank) to remove any sequences that did not match functionally relevant sequences.

[0058] The DNA synthesis supplier fragments the sequence from the customer's order into all possible overlapping fragments, the same size as those in the database. The fragments are translated into equivalent peptides using all possible reading frames. The fragments from the customer's order are then automatically compared to those in the database.

[0059] result

[0060] The synthesis supplier can screen all orders for fragments that perfectly match those from the prohibited list, with very few or no false positives corresponding to irrelevant sequences. The screening can be done fully automatically, avoiding the costs associated with human experts.

[0061] Example 2

[0062] A database of test sequences is constructed by randomly selecting fragments, optionally biased towards highly conserved regions, from lists of prohibited organisms and toxin genes (such as those from the US Federal Select Pharmacy Program (FSAP) and the Australian Group Treaty). The functional equivalents of these fragments are calculated using a predictive procedure or algorithm, and random numbers are included in the database. This database is pre-screened against known databases (such as GenBank) to remove any fragments that match functionally irrelevant sequences.

[0063] The DNA synthesis supplier fragments the sequence from the customer's order into all possible overlapping fragments, the same size as those in the database. The fragments are translated across all possible reading frames to produce equivalent peptides. The fragments from the customer's order are automatically compared to those in the database to detect orders that would produce functional equivalents of prohibited organisms or toxin genes.

[0064] result

[0065] The synthesis supplier can screen all orders for fragments that are functionally equivalent to those from the prohibited list, with very few or no false positives corresponding to functionally irrelevant sequences. Screening can be done fully automatically.

[0066] Example 3

[0067] A database of test sequences is constructed by randomly selecting fragments, optionally biased towards highly conserved regions, from lists of prohibited organisms and toxin genes (such as those from the US Federal Select Pharmacy Program (FSAP) and the Australian Group Treaty). The functional equivalents of these fragments are calculated using a predictive procedure or algorithm, and random numbers are included in the database. This database is pre-screened against known databases (such as GenBank) to remove any fragments that match functionally irrelevant sequences.

[0068] The DNA synthesis supplier assigns an information key to each customer interested in protecting their orders from industrial espionage. Customer orders are fragmented into all possible overlapping fragments of the same size as those in the database, translated across all possible reading frames to produce equivalent peptides, and all results are hashed using the key. The supplier performs a similar hashing on all sequences in the database. Fragments from customer orders are automatically compared to those in the database without sharing the customer orders themselves to detect orders that would produce functional equivalents of prohibited organisms or toxin genes.

[0069] result

[0070] Synthesis suppliers can screen all orders for fragments that are functionally equivalent to those from a prohibited list, with very few or no false positives corresponding to functionally irrelevant sequences. Screening can be done fully automatically, without requiring customers to provide their orders to the synthesis supplier in plaintext, thus protecting customers from industrial espionage.

[0071] Example 4

[0072] A database of test sequences is constructed by randomly selecting fragments, optionally biased towards highly conserved regions, from lists of prohibited organisms and toxin genes (such as those from the US Federal Select Pharmacy Program (FSAP) and the Australian Group Treaty). The functional equivalents of these fragments are calculated using a predictive procedure or algorithm, and random numbers are included in the database. This database is pre-screened against known databases (such as GenBank) to remove any fragments that match functionally irrelevant sequences.

[0073] The DNA sequencing provider fragments sequencing results from customer samples into all possible overlapping fragments, the same size as those in its database. The fragments are translated across all possible reading frames to produce equivalent peptides. Fragments from customer orders are automatically compared to those in the database to detect substances used by the customer that could produce functional equivalents of prohibited organisms or toxin genes.

[0074] result

[0075] Sequencing providers can screen all sequencing results for those functionally equivalent fragments from the prohibited list, with very few or no false positives corresponding to functionally irrelevant sequences. Screening can be done fully automated.

[0076] Example 5

[0077] A database of test sequences is constructed by randomly selecting fragments, optionally biased towards highly conserved regions, from lists of prohibited organisms and toxin genes (such as those from the US Federal Select Pharmacy Program (FSAP) and the Australian Group Treaty). The functional equivalents of these fragments are calculated using a predictive procedure or algorithm, and random numbers are included in the database. This database is pre-screened against known databases (such as GenBank) to remove any fragments that match functionally irrelevant sequences.

[0078] DNA design software vendors fragment the sequences input by clients into all possible overlapping fragments, the same size as those in their database. The fragments are translated across all possible reading frames to produce equivalent peptides. Fragments from client orders are automatically compared to those in the database to detect clients who may have unintentionally or intentionally designed modified constructs with functions equivalent to genes from prohibited organisms or toxins.

[0079] result

[0080] The design software vendor can filter all designs that are functionally equivalent to those from the prohibited list, with very few or no false positives corresponding to functionally irrelevant sequences. The filtering can be done fully automatically.

[0081] Example 6

[0082] An embodiment of the sequence screening method of the present invention was studied. The experiment tested random samples of 10,000 variants of the window PQSVECRPFVFGAGKPYEF (SEQ ID NO: 23) within gene PIII of phage M13, an exemplary virus that infects E. coli bacteria and is harmless to humans. M13 was used as a representative virus in the study. In this case, a library of variant sequences was generated, sequenced, and repeatedly infected to select for functionally preserved mutants. After each round of selection, the survivors were sequenced to quantify the frequency of variation for each variant. FUNTRP and BLOSUM62 classifiers were used to generate fitness estimates in arbitrary units, whereby, for the purpose of determining the true extent, a phage was considered sufficiently “harmful” if the ratio of its representative proportions measured in a large phage population before and after propagation in bacterial cultures exceeded a certain threshold. In other words, the classifiers attempted to name… The elements, and The actual situation is based on the fitness limit f 最小 It was established experimentally.

[0083] The genome of M13 virus infecting *E. coli* was evaluated using the prediction tool funtrp (for protein sequences) and nucleic acid conservation (for nucleic acid sequences in viral replication origins and packaging signals). Experimental data were obtained regarding the effectiveness of variants that replaced 19 amino acid windows within proteins and variants that replaced 42 base pairs within functional nucleic acid sequences. Fifteen 42- or 57-base-pair sequences were identified for experimental studies of packaging signals, positive and negative replication origins, and genes I, II, III, and IV. Oligonucleotide libraries containing 220,000 sequences at locations predicted by funtrp or by structural analysis (for nucleic acids) were constructed to assess the accuracy of variant prediction. These variant libraries were cloned into phagemids, which were plasmids containing copies of the M13 replication origin and (for proteins) the gene encoding the relevant protein from the M13 virus (see [link to phagemid]). Figure 8Helper plasmids containing M13 phages with replication origins and packaging signals disrupted by insertion of the p15a plasmid origin and kanamycin resistance gene were constructed. Each helper plasmid contained all protein-coding genes (for nucleic acid studies) or missing gene I, gene II, gene III, or gene IV. For example, a helper plasmid missing gene I could be supplemented by a phage plasmid library encoding a gene I variant to produce an M13 particle encoding the phage plasmid library instead of the helper plasmid. These were mixed with recipient *E. coli* cells, and recipient cells successfully infected carrying phage plasmids were selected to effectively select phage plasmids capable of supplementing the missing genes. The degree of enrichment or de-enrichment relative to the wild-type sequence corresponds to the variant's fitness in terms of viral production and infection.

[0084] A library of gene variant III with the amino acid sequence PQSVECRPFVFGAGKPYEF (SEQ ID NO: 23) was initially cloned into DH5α cells and sequenced using MiSeq (see [link to MiSeq]). Figure 9 The initial library diversity was measured (NGS point 1, initial library). These were then transformed into cells carrying helper plasmids with deleted gene III and sequenced again (NGS point 2, preselection). The resulting cells were cultured, and the M13 particles were purified and sequenced (NGS point 3, phage extrusion), subsequently mixed with acceptor cells carrying different antibiotic resistance markers. The resulting cells were cultured to select both phage particles and acceptor markers and sequenced (NGS point 4, post-selection). Selection was repeated twice more to obtain additional enrichment data (NGS points 5 and 6, post-selection and post-selection). The library sequencing coverage was approximately 40×, with 100% coverage for the two preselected samples.

[0085] The selection resulted in enrichment / de-enrichment of four orders of magnitude in each direction. Figure 10 This indicates that the variants were indeed selected through certain criteria. Fate analysis of all single-mutant variants for specific amino acid residues showed that alanine at position 13 allowed any mutation, while proline at position 1 did not. Figure 11 and 12 ).

[0086] A variant is considered fit if the ratio of the measured proportion of variants in a larger population before and after selection (NGS point 1 versus NGS point 4 or 6) exceeds a certain threshold of several thresholds. These distributions of fit and unfit sequences from the library are used as empirical datasets to evaluate predictions.

[0087] For predictive functions, each location is charged with a zero, medium, or high fitness cost. Figure 13FunTP analysis of the probability of substitutions was combined with a BLOSUM62 matrix defining common substitutions for amino acids at these positions to generate predictions of functional variants for all variants in the library. These were compared with empirical results from multiple threshold limits to construct ROC curves and evaluate the true positive and false positive rates of the combined funTP+BLOSUM62 classifier. Notably, the ROC curves for NGS points 1 to 4 and 1 to 6 were almost identical, indicating that only one round of selection is required.

[0088] Figure 14 A graph showing the ROC curves generated from the experimental data is presented. The graph represents the receiver operating characteristic (ROC) curves of a weak classifier based on the tools FUNTRP and BLOSUM62, as assessed based on the biologically fundamental fact data obtained from the experimental study. The ROC curves capture the trade-off between Type I (false positive) and Type II (false negative) errors for a yes-no classifier. The false positive rate (x-axis) is the proportion of variants not classified as fit, and the true positive rate (y-axis) is the proportion of fit variants correctly classified. The curves are parameterized by s∈(-∞, ∞), and the fitness threshold is expressed in arbitrary units; the classifier is used to distinguish between positive and negative based on its noisy fitness estimate. Figure 15 Is it to display the same classifier (e.g.) Figure 14 Another receiver operating characteristic (ROC) curve plot (NGS) is shown, comparing the enrichment / de-enrichment scores of NGS point 1 and NGS point 4. The similarity of the ROC curves indicates that repeated selection is unnecessary.

[0089] Similar evaluations can be performed on other classifiers and other variant libraries to improve predictions as needed. For nucleic acids, predictions can be combined with a conservation analysis at each location, along with structural analysis that calculates changes in the folding energy of the relevant RNA secondary structures due to mutations.

[0090] It is noteworthy that the ROC curve of funtrp+BLOSUM62 is sufficient to predict 90% of the functional sequences in the library, at the cost of half of the sequences being false positives. That is, given 10,000 functional sequences, the ROC curve shows that the classifier can predict 18,000 and successfully cover 9,000 of the 10,000. If such sequences are included in the database across multiple locations of harm, the probability of detection becomes very high, and considering the cost of including enough variants to have a chance to evade detection, the probability of the adversary obtaining functional sequences without detection becomes very low.

[0091] Example 7

[0092] The test sequence database is constructed by randomly selecting fragments from lists of prohibited organisms and toxin genes (e.g., those from the US Federal Select Agent Program (FSAP) and the Australia Group treaty), optionally biased towards highly conserved regions. Functional equivalents of these fragments are calculated using a prediction procedure or algorithm, and random numbers are included in the database. The database is pre-screened against known databases (e.g., GenBank) to remove any fragments that match functionally irrelevant sequences.

[0093] Opponents attempt to synthesize functional versions of banned genes or genomes. How can they determine their chances of success or failure?

[0094] The following describes an analysis of the screening method of the present invention. The term "secureDNA system" refers to one embodiment of the screening method system of the present invention. The system is used to identify "harmful" sequences and / or potential functional variants of harmful sequences. In the following description, the term "individual" means an organism, such as a virus, bacteria, or other organism. Nucleotide sequences from an individual are evaluated using the following methods to determine whether they are functional sequences, for example, whether they allow the organism to survive and replicate if contained in the organism. In this embodiment, the term "adversary" means a person or entity that is interested in synthesizing or has synthesized sequences considered harmful polynucleotide sequences. In the following description, the term "defender" means an operator of the system of the present invention who attempts to prevent unauthorized persons and entities from synthesizing or otherwise accessing dangerous polynucleotide sequences.

[0095] If the SecureDNA system prevents all adversaries from assembling sequences encoding functional biological hazards, then the system has successfully screened DNA. The most dangerous hazard variants are self-replicating agents, which are capable of spreading exponentially without human intervention. The functional sequence of such a replicator is defined as a DNA sequence with sufficient fitness to survive and replicate in a shared environment, thus becoming increasingly prevalent without human intervention (e.g., a new pandemic virus). Fitness can be formalized in several ways: the probability of an object surviving and reproducing, the expected number of offspring of an object, or any of these normalized for some relevant population. In any case, a real number representing a similar probability in [0,1] is a sufficient representation of fitness, and it can be assumed that there exists some minimum fitness f. 最小 Below the minimum fitness f 最小The reagent will die out. If the maximum fitness of all hazardous variants that can still be synthesized using SecureDNA is less than f... 最小 If so, the SecureDNA system is successful.

[0096] SecureDNA uses a Random Adversarial Threshold (RAT) screening method to search for hazardous fragments and potential functional variants. A variant is a window of DNA or amino acid sequence that differs from the wild-type sequence at one or more locations at the same locus (e.g., the sequence of a real reagent that can be found online). Each hazardous fragment consists of many loci, and any window within any locus allows for any variant. The conditional distribution F(v) is defined as the fitness or function of a hazardous fragment for a given variant v, where v is a triple (h, l, s). v h: hazard marker or index; l: window index within the genome; s v : The exact variant sequence. The fitness of the wild type at any locus F((h,l,s) h：l The definition is 1. This is because as long as F((h, l, s) is defined as 1. v ))>∈, sequence variants s v Typically, both hazards and windows are unique; with a slight overuse of symbols, one could say F(s) v = F(v). The total number of windows in the coding sequence of the hazard is denoted as N. Complex interactions between variants are possible, but it is assumed that in small fitness adjustments of the wild type, at least multiplicative complexes (i.e., fitness of hazard with multiple variants) are at most the product of individual variant fitness. Individuals using the SecureDNA system privately select windows to screen and select variants to include within each hazard using prediction software available in the art. Experiments were conducted using funtrp [Miller, M., et al., Nucleic Acids Research, 2019, Vol.47, No.21, e142] to combine the BLOSUM62 matrix of amino acid substitution probabilities [Eddy, S. Nat Biotechnol 22, 1035-1036 (2004)].

[0097] For a given RAT database The opponent's task is to select a set of variants V such that Empty, and

[0098]

[0099] This constitutes a failure of SecureDNA. In this embodiment, it is conservatively assumed that the opponent has an oracle capable of perfectly predicting the fitness of any given variant, i.e., the opponent knows F(v). It is worth noting that currently available methods for estimating F(v) are extremely poor, and therefore the information provided also includes some explanation for the significant inaccuracy of this estimate, which is a realistic condition for the evaluation.

[0100] 2. Breakthrough change approximation

[0101] The actual fitness distribution can only be measured empirically and partially; a given experiment will attempt to evaluate over a few million variants within a single window and then only for measurable biomolecules. This study allows for a rough estimate of the fitness distribution for the most basic and evolutionarily conserved windows: for example, a sequence may allow substitutions at nine of 19 amino acid residues with no more than a moderate fitness cost, allowing seven, seven, five, five, four, three, three, one, and one substitution residue, totaling 737,280 variants that do not completely disrupt harmful function for that particular window (F(v) > 0.5). This situation is conservatively approximated by assigning all of these a value of 1 and all remaining variants a value of 0.

[0102] A "breakthrough change" approximation F is introduced for the fitness distribution. b (v)

[0103]

[0104] in These are a group of variants that are functionally similar to the wild-type sequence. This similarity is good, for example, when an amino acid plays a crucial topological or affinity role in its protein, so only a small subset of substitutions results in a functional protein and harm. It is assumed that individuals selected for screening the key regions meet the criteria for this similarity.

[0105] The system involves selecting k distinct windows, i.e., k indexes w into the harmful genome. Let V w It is all variants (h,w,s) v The set of all functional variants in this window is... RAT database The coverage of window w is

[0106]

[0107] Assuming there are no preferences among the functional variants and that they all have effectively equal fitness, the probability that an opponent randomly chooses a functional variant not present in the database at all k positions is called the "escape" event, known as "E".

[0108]

[0109] The significant bound of the avoidance probability is

[0110]

[0111] From the arithmetic-geometric mean inequality. More roughly, it can also be based on the maximum coverage α of all k windows. 最大 Introducing boundaries:

[0112] P(E)≤1-α 最大 (3)

[0113] If we can even establish a strong coverage guarantee, starting from 3, we can rely on the maximum coverage provided by any window to constrain P(E). With F b (v) A defender with complete knowledge of the opponent's defenses may be able to completely cover the lowest... One or more windows, thereby achieving α 最大 A perfect defense with P(E) = 1 and P(E) = 0.

[0114] However, the defender does not possess F b (v) Complete knowledge. From the defender's perspective, predicting which variants are and are not. The members are uncertain, and even the degree of uncertainty in such a prediction is difficult to estimate. Given that only weak guarantees can be established on average coverage, it can be seen from (2) that one way to compensate is to include more windows, i.e., increase k.

[0115] A stronger bound is explored in Section 4.1 below, which also takes advantage of the adversary's lack of knowledge of the positions of the k windows. First, the following discussion involves the defender estimating F. b The impact of uncertainty in (v) is assumed to be known to the opponent regarding the defender's choice of k windows.

[0116] 3. Incorrect compromise

[0117] Assume the prediction is a trade-off between Type I and Type II errors, making each additional entry for w increasingly likely a false positive, as the defender attempts to cover a larger proportion using a noisy classifier. The receiver operating characteristic (ROC) curve of a classifier summarizes this trade-off, traditionally given by the following...

[0118]

[0119] Among them tp w (s) and fp w (s) represent the true positive rate and false positive rate for identifying functional variants, respectively, i.e., the ability to distinguish between them. The members are , and s is a threshold parameter indicating how aggressively we include potential functional variants.

[0120] ROC curves accurately capture the trade-off between Type I and Type II errors.

[0121] Selecting points on the curve based on the selection criteria, known as operation points, constitutes a specific compromise scheme, which can be selected in a principled manner.

[0122] There are many methods to quantify and optimize ROC curves. A useful example is to quantify the cost C. tp C fp C fn and C tn These are defined as the costs of true positive, false positive, false negative, and true negative test results in a game-theoretic sense, respectively. Then, assuming a convex and differentiable ROC curve (usually generated by fitting data), it can be obtained according to the tangent criterion at s = s w，opt The unique optimum on the selection curve [see England, WL, Medical Decision Making, 1988 vol. 8(2): 120-131, the contents of which are incorporated herein by reference in their entirety.]

[0123]

[0124] Where q is the "basic rate" of the functional sequences in the considered subset of the sequence space. That's indeed the case.

[0125]

[0126] Although considering This is almost a hollow lower bound relative to the sheer number of functional sequences. In fact, no adversary or defender will choose a variant outside a Hamming distance *r* until the variant becomes too different from the wild type to be effective. *r* is an empirical biological parameter. A potential expression of *q* might be...

[0127]

[0128] in In the set The volume of a Hamming sphere with an inner radius of r. The volume of the Hamming sphere can be understood as the size of the set of reasonable variants that can be imagined to operate a priori.

[0129] Once selected w，opt , covering α w Given from the following

[0130] α w=tp w (s w，opt ).

[0131] This method is attractive because it provides

[0132] 1. Methods for coherently integrating and comparing experimental data on actual fitness and predicted fitness for certain fitness estimation methods by constructing empirical ROC curves (e.g., based on next-generation sequencing (NGS) data from harmless virus populations in control experiments), and...

[0133] 2. Capture the coherent integration of interpretable cost parameters in the trade-offs that are at play.

[0134] This system strikes a balance between the probability of evading screening, the size of the RAT database, and the associated global rate of misclassifying random sequences as hazardous. The global false alarm rate, i.e., the probability of classifying a random sequence as hazardous, is...

[0135]

[0136] in It is the collection of all sequences with the same length as the window; a protein window of 19 amino acids has 20 [units / lengths]. 19 There are 4 elements, and the DNA window of 42 base pairs has 4 42 There are 10 elements. Since the total amount of DNA synthesized each year is increasing at an accelerating rate, g should be kept as low as possible.

[0137] In terms of the global false alarm rate due to random misclassification, any content in the database will incur the same cost, C. tp =C fp :=C p The cost of a true negative is zero (C). tn =0). The standard for the tangent in (4) becomes

[0138]

[0139] Furthermore, due to its relationship with g, C p Inversely proportional to |S|, where |S| is an exponent of the window length. The window is as long as possible, preventing the easy assembly of longer DNA sequences from short sequences that are unselectable due to their shorter length (approximately 50 base pairs), and is an inherent physical property of DNA. This constraint prevents the assembly of C... p The reason for driving to any low.

[0140] False negative C fn The cost has not yet been discussed. fnThe likelihood of adversaries exploiting false negatives to increase the expected availability of P(E) is likely the subject of detailed analysis. In particular, it depends on the database coverage and current size. Currently, it is considered an external parameter to examine its impact.

[0141] To establish an example, under the simplified assumption that all k windows have the same ROC, the subscript w is removed, and from Equation 1...

[0142] P(E)=(1-α) k .

[0143] This example demonstrates how the quality of the classifier captured by its ROC curve affects the optimal choice of parameters, especially k.

[0144] The instances used are for all This indicates that 5 out of 20 possible amino substitutions are effective at each of the 4 positions in each window. The maximum Hamming distance has been determined to be 6 before any further variations can be imagined to be effective. The volume of a Hamming sphere with a radius of 6 and a chord of length 19 in a 20-alphabet is...

[0145]

[0146] q≈4.8x10 -10 It is the ratio (Equation 5). If the value is set to C fn =10 8 C p In other words, the cost of ignoring the inclusion of functional variants in the database is 100 million times greater than including additional items in the database (imagine the scale of the impact of a successful harmful synthesis).

[0147]

[0148] Although we don't have the closed-form form or data of the classifier's ROC curve at hand, we can intuitively understand the relationship between the classifier's "quality" and the bound that can be placed on P(E). Qualitatively speaking, a "high-quality" classifier clearly distinguishes between functional and non-functional variants. It has an ROC curve that is steep near fp=0 and fp=1, reaching a high point at (fp,tp)=(0,1). The area under the ROC curve (AUC) is closer to 1. It may be within a range There is a slope. In contrast, the ROC curve of the "low-quality" classifier is closer to the line tp=fp, and its AUC is closer to 0.5, which means that its performance will not exceed the chance and may have a slope in the range [2 / 3, 3 / 2].

[0149] Assuming a high-quality classifier at fp(s) opt ) = 3 x 10-6 ;tp(s opt The target slope of 21 is reached at α = 0.95. The explanation is that by covering 0.0003% of the Hamming sphere around reasonable variants of the wild-type sequence, corresponding to a database size of approximately 1 million, it has achieved α = 95% coverage, given the defined cost C. p and C fn The specified balance would be the ideal compromise.

[0150] The target bound for the avoidance probability is set to P(E) = 0.001, ensuring that the adversary will inflict a functional harm only once in an average of 1000 complete orders. Using a high-quality classifier, the number of windows k that must be covered is...

[0151]

[0152] Assume that a low-quality classifier is actually available. This classifier has no s opt Make The explanation is that low-quality classifiers cannot be used to achieve the best trade-off between given cost balances. Instead, s was chosen. opt To provide a maximum actual database size of 10 million, corresponding to fp(s) opt ) = 3.2 x 10 -5 Because the ROC curve is close to a straight line tp(s) opt ) = fp(s opt True positive rate tp(s) opt It cannot be higher, for example, 6×10 4 To achieve the same boundary P(E) = 0.001, the number of windows k that must be covered is...

[0153]

[0154] This is impossible except for replicators with the largest genomes, and even then, such a database size would never be used.

[0155] The main benefits of the above application are:

[0156] For any dataset where the fitness of experimental measurements is compared with a given computational tool for predicting protein or DNA function, the ROC curves of classifiers used to approximate breakthrough changes can be readily obtained empirically, plotted, and analyzed.

[0157] • Explicit costs may be related to whether or not variants are included in the database, and these can set the ideal operating point for each classifier.

[0158] A weak classifier can be used, but it requires a larger window (larger k) to compensate for its poor true positive rate. k is a function of quality and is evaluated only by the ROC and cost settings.

[0159] There is a complex interplay between cost setting and optimal operating point. For example, the number of cryptographic operations that k performs directly affects C. p This directly translates to a DOPRF call, and the adversary's anticipated availability of a single false negative for increasing P(E) directly affects C. fn The interactions between these parameters are expected to be convex, and the solution can be handled using numerical methods.

[0160] • It has a minimum average classifier strength, which is necessary to appropriately limit the probability of escaping the selection due to cost constraints.

[0161] 4. Random Overlay Variants

[0162] When a defender approaches complete knowledge of F(V), it may deterministically select windows to protect because they require the fewest database entries to bind P(E), which may arbitrarily approach zero if most functional variants can be collected. Once these fully protectable windows are covered (if the selection is based solely on this criterion), it seems that including any other windows offers no benefit. An adversary with F(V) oracle knowledge knows this and can focus its attention solely on these areas, leveraging its superior fitness predictions to seek counterintuitive functional variants that are unlikely to be screened. Paradoxically, the fewer functional variants it has, the simpler the screening harm becomes, and the easier it is for an adversary with superior fitness estimations to evade screening, as long as the database construction focuses solely on deterministically covering certain windows.

[0163] A random defender strategy can be used to increase the expected work that an adversary with knowledge of the F(V) oracle must do up to an impossible degree by indeterminately selecting a window.

[0164] 4.1 Forces adversaries to modify more windows

[0165] This section describes how to strengthen the boundaries in Section 2 (above). To do this, it is observed that there are at most N windows in total, for which... There may be entries in it. On the other hand, due to practical constraints, it may be desirable to add modifications to k of them to it. Section 2 makes an implicit assumption that the adversary actually knows which k windows they must modify, but in practice, the adversary does not actually know this.

[0166] One approach is to represent the set of variants sent by the opponent as (where it is assumed that the windows do not overlap). When compared with the target threat to be defended against, the adversary will modify the value of l. The first observation is that l≥k is necessary: if an adversary modifies strictly fewer than k windows, but the original sequence of each such window is in D, the adversary will always be caught.

[0167] Next, consider an adversary that modifies all N windows. Its chance of successfully passing the test with a "fit" sequence is where is the fitness of the actual sequence according to our previously mentioned fitness function.

[0168] is the probability of not being caught by RAT.

[0169] Setting l=N, then can be bounded exactly as in Section 2. But in addition, the success of an adversary lacking a fitness oracle is affected by the fitness term, which is extremely likely to be 0 if the entire sequence must be modified.

[0170] To establish the bound for k≤l<N, let A t denote the event that the adversary chooses l modifications such that all k windows protected by the adversary are included. Furthermore, still let P(F) be the probability of not being detected in any of the k windows, and let E be the corresponding event. Since the wild-type sequence is definitely in the database, the adversary must as a result have passed at least all k tests, and simultaneously identified the correct k out of N windows using l modifications from the wild type. Therefore and the success probability of the adversary becomes

[0171]

[0172] where is the sequence with l modifications. The standard method for determining the upper bound of the adversary's success probability is to find the local maximum with respect to l, and then select k appropriately, which requires that the above function is differentiable [in particular . Explicit bounds for P(E) have been given in Section 2, so it is necessary to analyze the other terms.

[0173] In summary, Example 7 provides a mathematical evaluation of the extreme challenges faced by even well-prepared adversaries attempting to synthesize sequences protected by the system described in this invention. This evaluation provides insight into the effectiveness of the screening method of this invention. The greater the proportion of functional variants of a particular window present in the database, the lower the probability of evading detection. The more protected windows there are, the lower the probability of evading detection. Given that adversaries can perfectly predict the functionality of the resulting sequences—a separate and more difficult problem compared to the unresolved issue of perfectly predicting the fitness of variants within a particular window—it will be difficult for them to evade screening. These results suggest that, given enough sequences in the database, real-world adversaries risk including mutations that render their sequences nonfunctional, and the chances of adversary success are negligible.

[0174] Equivalent scheme

[0175] While several embodiments of the invention have been described and illustrated herein, those skilled in the art will readily conceive of a variety of other ways and / or structures for performing the functions and / or obtaining the results and / or one or more advantages described herein, and each such variation and / or modification is considered to be within the scope of the invention. More generally, those skilled in the art will readily understand that all parameters, sizes, materials, and configurations described herein are intended to be exemplary, and actual parameters, sizes, materials, and / or configurations will depend on one or more specific applications using the teachings of the invention. Those skilled in the art will recognize, or can determine, many equivalents of the specific embodiments of the invention described herein using only conventional experimentation. Therefore, it should be understood that the foregoing embodiments are given by way of example only, and that the invention can be practiced in ways other than those specifically described and claimed within the scope of the appended claims and their equivalents. The invention relates to each individual feature, system, article, material, and / or method described herein. Furthermore, any combination of two or more such features, systems, articles, materials, and / or methods is included within the scope of the invention if such features, systems, articles, materials, and / or methods are not inconsistent with each other.

[0176] All definitions defined and used herein shall be construed as taking precedence over dictionary definitions, definitions incorporated by reference in other documents, and / or the general meaning of the defined terms. Unless expressly stated otherwise, nouns without quantifiers as used herein in the specification and claims shall be construed as meaning "at least one of them".

[0177] The phrase “and / or” as used herein in the specification and claims should be understood to mean “one or both” of the elements so connected, i.e., the elements coexist in some cases and exist separately in others. Unless explicitly stated otherwise, other elements may optionally exist besides those explicitly indicated by the “and / or” clause, whether related to or unrelated to those explicitly indicated elements.

[0178] All references, patents and patent applications and publications cited or referenced in this application are incorporated herein by reference in their entirety.

Claims

1. A method including a computer-executed means for evaluating a biological sequence capable of performing a pre-selected function, comprising: (a) Pre-selecting biomolecules, wherein the pre-selected biomolecules are capable of achieving the intended function; (b) Prepare a test sequence database containing multiple sequence fragments of preselected biomolecules, wherein the multiple sequence fragments have a predetermined length; (c) Identify one or more sequence fragments of the preselected biomolecule in the test sequence database, wherein one or more of the sequence fragments match one or more sequence fragments of a second biomolecule having a biological function unrelated to the intended function of the preselected biomolecule, and remove the identified one or more sequence fragments from the test sequence database. (d) In the test sequence database, the sequence fragments of the test biomolecule are fragmented into lengths equal to the predetermined lengths of multiple sequence fragments of the pre-selected biomolecule; (e) Detecting the presence or absence of a sequence match between at least one fragment of the fragmented test biomolecule and at least one of multiple sequence fragments of the pre-selected biomolecule in the test sequence database, and (f) functions in response to the detection in (e), wherein the detection in (e) provides an assessment of the test biomolecule, wherein the function includes one or more of the following: preventing the synthesis of the test biomolecule, allowing the synthesis of the test biomolecule, sequencing one or more polynucleotide molecules, DNA sequencing, DNA molecule design, peptide sequencing, and other sequence identification steps.

2. The method of claim 1, wherein if the presence of a sequence match is detected in (e), the action in (f) includes preventing the synthesis of the test biomolecule.

3. The method of claim 1, wherein the means of preparing the test sequence database includes: (i) Screening multiple sequence fragments of a preselected biomolecule against at least one control sequence database, wherein the at least one control sequence database contains multiple control sequence fragments of at least one molecule, the at least one molecule being capable of achieving a purpose function unrelated to the purpose function of the preselected biomolecule; (ii) Identify the match between a sequence fragment from a plurality of sequence fragments of the preselected biomolecule and a control sequence fragment from at least one control sequence database; as well as (iii) Remove the sequence fragments of the preselected biomolecules that were identified as matching the control sequence fragment from the test sequence database.

4. The method of claim 1, wherein the preselected biomolecule is a polynucleotide.

5. The method of claim 4, wherein the sequence of the preselected biomolecule is the full-length nucleic acid sequence of the polynucleotide or a portion thereof.

6. The method of claim 5, wherein the full-length nucleic acid sequence encodes a protein.

7. The method of any one of claims 4 to 6, wherein the predetermined length of the plurality of sequence fragments of the preselected polynucleotides is 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200 or more nucleotides.

8. The method of any one of claims 1 to 4, wherein the preselected biomolecule comprises a polypeptide.

9. The method of claim 8, wherein the sequence of the preselected biomolecule is the full-length amino acid sequence of the polypeptide, or a portion thereof.

10. The method of claim 8, wherein the predetermined length of the plurality of sequence fragments of the polypeptide is 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 2, 93, 94, 95 or more amino acids.

11. The method of claim 1, wherein the plurality of sequence fragments of the preselected biomolecule: (1) comprises all of the possible fragments of the preselected biomolecule capable of achieving the intended function, and (2) does not contain sequences present in biomolecules capable of achieving functions unrelated to the intended function of the preselected biomolecule.

12. The method of claim 3, wherein the at least one control sequence database comprises multiple control sequence fragments of at least one molecule, wherein the at least one molecule cannot achieve the intended function of the preselected biomolecule.

13. The method of claim 1, wherein the predetermined length is the same for all of the plurality of sequence fragments of the preselected biomolecule.

14. The method of claim 1, wherein the predetermined length of the plurality of sequence fragments of the preselected biomolecule includes more than one length.

15. The method of claim 1, wherein the test sequence database comprises one or more sequence fragments randomly or pseudo-randomly selected from molecular sequences known to be capable of achieving the intended function of the preselected biomolecule.

16. The method of claim 15, wherein one or more of the randomly or pseudo-randomly selected sequence fragments are biased toward sequence regions that have higher homology to functional or phylogenetic sequences.

17. The method of claim 1, wherein the test sequence database further comprises sequences that are functional equivalents of multiple sequence fragments of the preselected biomolecules.

18. The method of claim 17, wherein the means for identifying the functional equivalent includes computer means.

19. The method of claim 17, wherein the means for identifying the functional equivalent includes experimental means.

20. The method of claim 18, wherein the computer means for identifying the functional equivalents includes using a classifier based on experimental data to evaluate the accuracy of the computer means.

21. The method of any one of claims 17 to 19, wherein the functional equivalents included in the test sequence database comprise a minimum number of sequences calculated to achieve a predetermined likelihood of successfully preventing the test sequences from escaping detection.

22. The method of any one of claims 17 to 19, wherein selecting functional equivalents included in the test sequence database comprises a random selection method or a pseudo-random selection method.

23. The method of claim 1, wherein the identity of all sequence fragments of the test sequence database and one or both of the test biomolecules is protected.

24. The method of claim 23, further comprising using an unintentional multi-party server system to apply a cryptographic hash function, said cryptographic hash function deterministically mapping sequence data to a fixed-size bit string using a one-way function.

25. The method of claim 24, wherein the application of the cryptographic hash function cannot be reversed without an exhaustive search of all possible sequences input into the test sequence database.

26. The method of claim 25, wherein applying the cryptographic hash function further comprises using one or more information keys, which must be accessed to attempt the exhaustive search.

27. The method of claim 26, wherein applying the cryptographic hash function requires keys from multiple independent sources, which must cooperate to compute the hash without any server gaining access to the sequence data.

28. The method of claim 27, wherein the plurality of independent information sources includes independent computer servers.

29. The method of claim 1, further comprising dividing the test sequence database into two or more partial test sequence databases, and using one of the two or more partial test sequence databases in (e) to detect the presence or absence of sequence matching.

30. The method of claim 29, further comprising: If a sequence match is detected between one or more fragments of the test biomolecule in one of the two or more partial test sequence databases, the presence or absence of one or more fragments of the test biomolecule is detected in the other of the two or more partial test sequence databases.

31. A method for identifying biological sequences capable of performing a pre-selected function, comprising: (a) Pre-selecting biomolecules, wherein the pre-selected biomolecules are capable of achieving the intended function; (b) Preparing a test sequence database comprising multiple sequence fragments of a preselected biomolecule, wherein the multiple sequence fragments have a predetermined length, wherein preparing the test sequence database includes: (i) Screening multiple sequence fragments of a preselected biomolecule against at least one control sequence database, wherein the at least one control sequence database contains multiple control sequence fragments of at least one molecule, the at least one molecule being capable of achieving a purpose function unrelated to the purpose function of the preselected biomolecule; (ii) Identifying the match between a sequence fragment from a plurality of preselected biomolecules and a control sequence fragment in at least one control sequence database; and (iii) Remove the sequence fragments of the preselected biomolecules that were identified as matching the control sequence fragment from the test sequence database; (c) Fragmenting the sequence of the test biomolecule in the test sequence database into segments of a predetermined length equal to the length of multiple sequence segments of the pre-selected biomolecule; and (d) Detect the presence or absence of a sequence match between at least one fragment of the fragmented test biomolecule and at least one of the multiple sequence fragments of the preselected biomolecule in the test sequence database.

32. The method of claim 31, wherein the preselected biomolecule is a polynucleotide.

33. The method of claim 32, wherein the sequence of the preselected biomolecule is the full-length nucleic acid sequence of the polynucleotide or a portion thereof.

34. The method of claim 33, wherein the full-length nucleic acid encodes a protein.

35. The method of claim 32, wherein the predetermined length of the plurality of sequence fragments of the preselected polynucleotides is 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200 or more nucleotides.

36. The method of claim 31, wherein the preselected biomolecule comprises a polypeptide.

37. The method of claim 36, wherein the sequence of the preselected biomolecule is the full-length amino acid sequence of the polypeptide, or a portion thereof.

38. The method of claim 36 or 37, wherein the predetermined length of the plurality of sequence fragments of the preselected polypeptide is 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95 or more amino acids.

39. The method of claim 31, wherein the plurality of sequence fragments of the preselected biomolecule: (1) comprises all of the possible fragments of the preselected biomolecule capable of achieving the intended function, and (2) does not contain sequences found in biomolecules capable of achieving functions unrelated to the intended function of the preselected biomolecule.

40. The method of claim 31, wherein the at least one control sequence database comprises multiple control sequence fragments of at least one molecule, the at least one molecule being unable to achieve the intended function of the preselected biomolecule.

41. The method of claim 31, wherein the predetermined length is the same for all of the plurality of sequence fragments of the preselected biomolecule.

42. The method of claim 31, wherein the predetermined length of the plurality of sequence fragments of the preselected biomolecule includes more than one length.

43. The method of claim 31, wherein the test sequence database comprises one or more sequence fragments randomly or pseudo-randomly selected from molecular sequences known to be capable of achieving the intended function of the preselected biomolecule.

44. The method of claim 43, wherein one or more of the randomly or pseudo-randomly selected sequence segments are biased toward conservative regions.

45. The method of claim 31, wherein the test sequence database further comprises sequences that are functional equivalents of multiple sequence fragments of the preselected biomolecules.

46. ​​The method of claim 31, wherein the identity of all sequence fragments of the test sequence database and one or both of the test biomolecules is protected.

47. The method of claim 46, further comprising applying a cryptographic hash function, wherein the cryptographic hash function uses a one-way function to deterministically map sequence data to a fixed-size bit string.

48. The method of claim 47, wherein the applied cryptographic hash function cannot be reversed without an exhaustive search of all possible sequences input into the test sequence database.

49. The method of claim 48, wherein the applied cryptographic hash function further comprises using one or more information keys, which must be accessed to attempt the exhaustive search.

50. The method of claim 47, wherein the applied cryptographic hash function requires keys from multiple independent sources that must cooperate to compute the hash without any server gaining access to the sequence data.

51. The method of claim 50, wherein the plurality of independent information sources includes independent computer servers.

52. The method of claim 31, further comprising dividing the test sequence database into two or more partial test sequence databases, and using one of the two or more partial test sequence databases in (d) to detect the presence or absence of sequence matching.

53. The method of claim 52, further comprising: If a sequence match is detected between one or more fragments of the test biomolecule in one of the two or more partial test sequence databases, the presence or absence of one or more fragments of the test biomolecule is detected in the other of the two or more partial test sequence databases.

54. The method of claim 31, further comprising acting in response to detection in (d), wherein acting includes one or more of the following: preventing the synthesis of the test biomolecule, allowing the synthesis of the test biomolecule, sequencing one or more polynucleotide molecules, DNA sequencing, DNA molecule design, determining the amino acid sequence of a polypeptide, and additional sequence identification steps.

55. The method of claim 31, further comprising acting in response to detection in (d), wherein if the presence of a sequence match is detected in (d), the acting includes preventing the synthesis of the test biomolecule.

56. The method of claim 31, further comprising identifying one or more sequence fragments of a plurality of preselected biomolecules in a test sequence database, wherein one or more of the plurality of sequence fragments respectively match one or more sequence fragments of a second biomolecule having a biological function unrelated to the intended function of the preselected biomolecule, and removing the identified one or more sequence fragments from the test sequence database.

57. A test sequence database, which is prepared by any one of claims 1 to 30.

58. A method for evaluating biological sequences, using the test sequence database of claim 57.

59. The method of claim 58, wherein the evaluation includes determining whether to allow or prevent the synthesis of the evaluated biological sequence.

60. A test sequence database, prepared by any one of claims 31 to 56.

61. A method for evaluating biological sequences, using the test sequence database of claim 60.

62. The method of claim 61, wherein the evaluation includes determining whether to allow or prevent the synthesis of the evaluated biological sequence.

Images