KR20120083521A - Computer-implemented biological sequence identifier system and method - Google Patents

Abstract

The present invention comprises the steps of transferring the reference sequence to the taxonomic data to produce a taxonomic result; And recording the taxonomy identification based on the taxonomic results. The reference sequence is the output of a genetic database query that returns a score for each reference sequence. The present invention comprises the steps of converting a base call in a predetermined position list in the biological sequence to N; And a method for processing a biological sequence obtained from the analysis by determining a ratio of single base polymorphisms in the biological sequence relative to the reference sequence. Each entry in a given location list represents the ability of the material to hybridize to the microarray used to generate the biological sequence. The substance is not the nucleic acid of the target pathogen.

Description

Computer-implemented biological sequence identifier system and method {COMPUTER-IMPLEMENTED BIOLOGICAL SEQUENCE IDENTIFIER SYSTEM AND METHOD}

Related application

This application is filed on June 16, 2005 in US Provisional Patent Application No. 60/691,768; US Provisional Patent Application No. 60/735,876, filed November 14, 2005; US Provisional Patent Application No. 60/735,824, filed Nov. 14, 2005; US Provisional Patent Application No. 60/743,639, filed May 22, 2006; US Provisional Patent Application No. 11/422,425, filed June 6, 2006; And US Provisional Patent Application No. 11/422,431 filed on June 6, 2006. This application is filed on July 2, 2005, filed in US patent application Ser. No. 11/177,647; US patent application Ser. No. 11/177,646, filed Jul. 2, 2005; And US patent application Ser. No. 11/268,373, filed Nov. 7, 2005. These regular applications are described in US Provisional Patent Application No. 60/590,931, filed July 2, 2004; US Provisional Patent Application No. 60/609,918, filed Sep. 15, 2004; US Provisional Patent Application No. 60/626,500, filed November 5, 2004; US Provisional Patent Application No. 60/631,437, filed Nov. 29, 2004; And US Provisional Patent Application No. 60/631,460 filed on November 29, 2004.

The present invention relates generally to methods of processing biological sequences.

For both the surveillance and diagnostic fields, micro-scale pathogen identification and near neighbor identification are important, so an assay monitoring at this very specific level is desirable in a clinical setting (1-3). In order to successfully use any method based on DNA or RNA detection, the assays must be linked to a number of databases of nucleic acid sequence information for analysis design and interpretation of the original data to ensure the desired information provision. . Some well-established techniques, such as real-time PCR, use short, distinct stretches of sequenced genomes to provide good specificity (4). These techniques can provide micro-scale identification of several genetically close organisms by selecting a sufficient number of fragments. However, these selected fragments, which are specific in the initial selection process, are later identified as often less specific as more organisms are sequenced. This is particularly problematic for pathogens belonging to families with high mutation rates and for pathogens with relatively non-contiguous pathogens identified. In addition, real-time PCR cannot detect the presence of new important mutations or analyze nucleotide sequence details. Similarly, the method of obtaining pathogen identification in other detection techniques has been improved, but suffers from some or all of the problems using PCR (5-6).

The high-density rearrangement microarray can produce ^{variable length fragments of 10 2} -10 ⁵ base pairs (bp) with direct sequence information. They have been used successfully to detect single nucleotide polymorphisms (SNPs) and genetic variations from viral, bacterial, and eukaryotic genomes (9-16). Their use in SNP detection has clearly established its ability to provide reliable quality sequence information. In most cases, microarrays are designed to study a limited number of genetically-like target pathogens, and in many cases, detection methods relied solely on recognizing hybridization patterns for identification (12, 14, 15). , 17, 18). Using the sequential sequencing power of rearrangement microarrays required for SNP detection, rearrangement has been recently successfully adapted for pathogen identification of many bacterial and viral pathogens by using different approaches, while closely related pathogens can be identified. It allowed microscopic identification and tracking of mutations in targeted pathogens (19-21). The new methodology differed from previous studies by using the analyzed base as a query for DNA database similarity investigation to identify the species and variants that would most likely match the base call from the observed hybridization. The system was able to test 26 pathogens simultaneously and detect the presence of multiple pathogens. A software program, REsequencing Pathogen Identifier (REPI), was used to simplify data analysis by performing similarity studies of genetic databases using BLAST (Basic Local Alignment Search Tool). (22). The REPI program used the BLAST default setting, meaning hybridization only when the expected value, i.e. the amount calculated by the BLAST program indicating the likelihood that the identified sequence matches will occur by random chance in the database is ^{less than 10 -9.} It will carry the sequence. This screened all cases with insufficient signal transduction, but the final decision as to which pathogens were detected and to what extent could be identified required manual examination of the returned results. This method successfully enabled strain identification of Flu A and B samples and micro-identification of various adenoviruses, consistent with the conventional sampling results (19, 20). Two important advantages of this research method were that information is always retrieved at the most detailed level possible, and that the information can still recognize organisms with the latest mutations. In addition, the study method maintained the specificity well, as it did not depend on the peculiarity of the short sequence, which was constantly encroached when more organisms were sequenced.

While useful, the analysis method has some drawbacks, namely, it is time consuming, not suitable for maximizing sensitivity, produces complex results, is suitable only for professionals, and contains redundant or redundant information. This process was time consuming because only the initial screening was handled automatically, while the remaining steps required manual interpretation before terminating the detection analysis. Since a simple criterion (expected definite range 10 ^-9 ) and non-optimized BLAST parameters are used to consider the detected pathogen, the REPI algorithm provides a list of candidate organisms, but fails to draw a final simple conclusion or I couldn't relate the results to another. Instead, a manual method was used for the final decision, but the REPI programs all provided the use of public nucleic acid databases containing similar results and redundant entries, thus providing users with a large amount of useless data. In addition, with a passive method, it was impossible to establish that the developed algorithm is generally applicable to any organism provided with sequence information from which the nucleic acid base was determined.

One method of the present invention includes the steps of generating a plurality of taxonomic results by transmitting a plurality of reference sequences to a taxonomic database as a query, and recording the taxonomic identification based on the taxonomic results. Includes. The reference sequences are the output of a genetic database query that carries the Score for each reference sequence.

Another method of the present invention for processing a biological sequence obtained from the analysis comprises the steps of converting a base call located on a list of positions in the biological sequence to N, and determining the ratio of a single nucleotide polymorphism in the biological sequence relative to the reference sequence And determining. Each entry in a given list of locations represents the ability of a material to hybridize to the microarray used to generate the biological sequence. The substance is not the nucleic acid of the target pathogen.

1 is a schematic diagram of an algorithm showing the relationship between three main tasks and the logic of subtasks related to the task. Task I, after performing filtering and subsequencing, determines which database record the original sequence is most similar to. Task II determines whether circular sequence identification supports identification of common organisms. Task III makes the final investigation and determination of organisms detected from microarray data. ProSeq: circular sequence; SubSeq: partial sequence; HybSeq: hybridization sequence.
2 is a detailed schematic diagram of the filtering sub-task of task I. For each ProSeq, after masking the primer regions as N (ambiguous) calls, UniRate was calculated from HybSeq. For ProSeq that passed the UniRate requirement, we tried a revised sliding window algorithm to increase SubSeq, which can be used as a query to BLAST. The substance of SubSeq that was successfully increased (starting position and length in ProSeq) was placed in a file for batch query through BLAST.
3 is a detailed schematic diagram of the subtasks of Task I involved in the identification of organisms for individual SubSeq. Each SubSeq sent to BLAST returned a list of possible matches contained in the retrieved Return array to find the best bit score/expected pair (MaxScore). If MaxScore was ^{greater than MIN(10 -6} ), all returns with the best score were sorted into a new array Rank1. The detailed determination process was described in the Methods section herein and then the organisms of SubSeq were identified.
Fig. 4 is a schematic diagram of a subtask of task I for determining an organism determined for ProSeq based on the results confirmed for SubSeq. All SubSeqs of a particular ProSeq are compared to each other to determine the SubSeq with the two best scores. When there is a single SubSeq or there is one with a score that is much better than the others, the ProSeq inherits the properties of the SubSeq. Otherwise, the common taxonomic class was determined as described in the patent specification.
5 is an alignment diagram of influenza A NA1 ProSeq and A/Weiss/43, A/Puerto Rico/8/34 strains. Also shown are the crude and filtered hybridization chip results of A/Puerto Rico/8/34. ^{* Indicates} a perfectly matched sequence.

A more complete understanding of the present invention is readily obtained by referring to the description of the following example embodiments and the accompanying drawings.

In the following description, for purposes of illustration without limitation, specific detailed descriptions are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced in other embodiments departing from these specific detailed descriptions. In other instances, detailed descriptions of well-known methods and devices have been omitted so as not to obscure the technical content of the present invention with unnecessary detailed descriptions.

As used herein, the term “sequence” refers to a nucleic acid sequence, such as DNA or RNA, or a protein sequence. As used herein, “base” and “base call” can mean a nucleotide base or an amino acid. As used herein, the term “taxonomic” can mean any level or class of identification of a pathogen, including, but not limited to, genus, species, strains, and subspecies. As used herein, the term “record” may include transmitting signals from one system to another and generating any form of human-readable report. All of the disclosed methods can be computer-implemented in an apparatus having means for performing the method.

A new software specialized system that can successfully use base sequence information determined from Affymetrix rearrangement microarrays designed to provide a simple list of detected organisms, Computer-Implemented Biological Sequence Identifier system) 2.0 (CIBSI 2.0) is disclosed. This algorithm addresses the shortcomings of prior methods by incorporating new features to fully automate pathogen identification. This single program, with improved sensitivity, can make accurate decisions for all 26 pathogens contained in the RPM v1 microarray, whether detected alone or in combination (19, 20, 23). ). This program is currently applied to rearranged microarrays, but a methodology developed because only the first part of the algorithm handles microarray-specific problems while the rest of the algorithm processes sequences suitable for use as queries by the BLAST algorithm. Becomes generally applicable. In developing a general identification algorithm, problems peculiar to rearrangement microarrays that complicate their use were identified and analyzed. Since we have automated the overall method of determining what is detected, it is easy to test that the rules used to identify are correct and applicable to any pathogen. Using this efficient program, an analysis based on rearrangement can provide a competitive way to test multiple possible pathogens simultaneously, while providing an output that can be interpreted by non-experts.

Amplification, hybridization, and sequencing

Detailed descriptions and experimental methods of RPM v1 microarray design were discussed in prior literature (19, 20, 23). Experimental microarray data used in the assays of the present invention were obtained using various purification templates and clinical samples using any of a variety of amplification instruments. GCOS software v1.3 (manufactured by Affymetrix, Santa Clara, CA) was used to align and scan the hybridized microarrays to determine the strength of each probe in all probe sets. The base call was constructed based on the intensity data of each probe set using GDAS v3.0.2.8 software (manufactured by Animatrix, Santa Clara, CA) that implements the ABACUS algorithm (11). The sequences were presented in FASTA format for later analysis steps.

The rearrangement microarray (RPM v1) is based on ProSeq without relying on a predetermined hybridization pattern, with a sequence of 20 common respiratory pathogens known to cause febrile respiratory disease and 6 CDC Category A biothreat pathogens. Designed in advance for classification and detection. Approximately 4000 RPM v1 experiments performed using various amplification techniques, single and multiple pathogen targets, purified nucleic acids and clinical samples were investigated to develop pathogen identification algorithms. The results of using this algorithm for clinical samples, identified pathogens and purified nucleic acids are discussed in detail in other studies (19, 20, 23). In all cases, the algorithm accurately identified organisms at the species or strain level, depending on the length of ProSeq shown in RPM v1. Several specific examples will be discussed to illustrate how the algorithm performs under various conditions.

The CIBSI 2.0 program handled three tiers of tasks (Figure 1): (I) determine which database record the detected organism is most similar to, and (II) identification from separate targets identifies common organisms. Support, and (III) whether the detected organism belongs to a target set specifically designed for the assay to detect or is associated with a close genetic contiguity. Target pathogens are organisms specifically designed for the assay to detect. As used herein, a set of probes representing a reference sequence selected from a target pathogen genome is referred to as a circular sequence or simply “ProSeq”. The set of determined bases resulting from hybridization of genomic material to ProSeq is referred to as the hybridization sequence or “HybSeq”. HybSeq is divided into possible subsequences or “SubSeq”. Part of the algorithm handles the identification of organisms based on ProSeq and includes three steps: first filtering individual HybSeqs with SubSeq suitable for sequence similarity comparison, querying the database for individual SubSeqs, and BLAST for each SubSeq. It is processed as a taxonomic comparison step of returns. In the next step, ProSeqs were compared to determine if they supported the same identified organism. In the final step, the detected organisms were compared to the list of target pathogens for which the assay was designed to determine if any organisms were detected as positive. The level of identification supported by a particular sample was automatically determined.

Filtering

The first filtering algorithm, i.e., REsequencing Pathogen Identifier (REPI), was first developed (20), and the general concept, including revision, was incorporated into the current (automated detection) algorithm used in the CIBSI 2.0 program. Filtering and partial sequencing were used to remove potential bias caused by reference sequence selection and by other sources (primers), as well as to separate HybSeq into fragments that were significant for faster exploration. This was the first sub-task of task I in FIG. 1 and is schematically illustrated in detail in FIG. 2. When PCR amplification was used, the microarray was hybridized only in the presence of the primers to determine where the primers caused hybridization. All portions of the hybridized ProSeq using primers were masked as N calls so that HybSeq did not contain biased information. For each ProSeq, the ratio of SNP to the total number of unique base calls, UniRate, was calculated from HybSeq. In order to remove HybSeq with insufficient hybridization when UniRate is ≥ 20% (SNP threshold), its ProSeq is considered negative for target organism detection. UniRate 20% indicated an average of 5 SNPs per 25 bp. For this frequency difference between the target pathogen-like organism and the reference sequence on which ProSeq is based, it is impractical to expect a significant specific hybridization of the 25 bp probe. This will terminate the filtering subtask and return to the task I loop, which will examine the next ProSeq. In the case of ProSeq with a ratio of <20%, a more detailed investigation was performed. At each position of HybSeq, a revised sliding window algorithm 20 was tried to increase SubSeq, which can be used as a query for BLAST. Initially, the first 20 bases (initial length) after a certain position were investigated as potential partial sequences. If less than 60% of this base was ambiguous (N), SubSeq entered the extension phase. SubSeq is 1 base at a time until the total content of unique base calls is reduced to 40% or less (specific base call threshold), or if the sliding window containing the final 21 bases has less than 4 unique base calls. Was extended. This differs from the REPI algorithm, in which only 20 bases sliding windows are used, and if less than 40% of the window content is a unique base call, the increase in SubSeq stops. At this point, the N calls trailing and traversing the SubSeq were removed. At least 1 position with 7 consecutive unique base calls was required to meet the word size parameter of BLAST and maintain SubSeq for further analysis. SubSeq greater than 100 bases was received. For acceptance, SubSeq of ≤ 30 bases required at least 95% specific base call (not "N"). For SubSeq with 30-100 bases, partial sequence acceptance requires at least VARI(("SubSeq length"-30)*0.2857+70)% unique bases. For SubSeq> 80 bases, the BLAST word size parameter was modified to 11 if it contained at least 11 consecutive bases. The entity of the successfully increased SubSeq (starting position and length in ProSeq) was placed in an entry in the SubSeq array that holds information related to each SubSeq. These entities and SubSeq were placed in a file for batch query through BLAST. This procedure was repeated by continuing from the end of the previous successful SubSeq, or, if unsuccessful, by continuing from the point where the window was initially increased to the end of the HybSeq. At the end, the algorithm returned to the task I loop and performed the BLAST subtask.

Database query

The BLAST subtask conducted a batch similarity investigation of the database using SubSeq as a query. The BLAST program used was NCBI Blastall -p blastn version 2.12 with a limited set of parameters. In the seeding step, the low-complexity region was shielded to speed up the query, but the actual scoring included the low-complexity repetition. The entire nucleotide database from NCBI obtained on February 7, 2006 was used as the reference database. (Note that during development, an initial image of the database was used, but all experiments were re-run with the algorithm described with the image of the database obtained on this date). The default gap penalty and nucleotide match score were used. The nucleotide mismatch penalty, -q, and parameters were set to -1 different from the default. The results of an arbitrary BLAST query with an expected value of <0.0001 were returned from the blastall program in tabular format. Information on each return (bit score, expected value, mismatch, match length) was placed in Return{hash key}{info} using the SubSeq entity as a hash key for further analysis.

Taxonomic-based pathogen identification for ProSeq from SubSeq

The next subtask of task I to be performed is the determination of the SubSeq() state, and is shown in FIG. 3. In order to present simple data and to facilitate the decision process, information for all SubSeqs was summarized by two parameters. "Identified organism" refers to the taxonomic class of organisms and "organism specificity" refers to the quality of the organism identification. The components in the return hash were examined and ranked by the score array for each distinct SubSeq() of ProSeq. The score array contained a pair of parameters with a fixed relationship to a given database, namely the bit score and the expected value. Sometimes it was appropriate to use a ranking score to describe the size of the database (expected value) or (bit score). The components in the return hash could have the same score, and all components with the highest bit score/lowest expected value (MaxScore) were held in a separate array Rank1. All taxonomic classifications of all components in Rank1 were retrieved from the NCBI taxonomic database, which was also obtained on February 7, 2006 (see note above). If the MaxScore expectation was ^{greater than MAX (currently 10 -6} ), SubSeq() retained both its identified organism and organism-specific information updated to null. If MaxScore was small enough, the bounces placed in Rank1 were examined. When Rank1 contains a single component, the organism specificity of SeqUniqu was assigned to SubSeq. If Rank1 contains multiple components, SubSeq is assigned, and TaxUnique's organism specificity is assigned if all transports belong to the same taxonomic class. Otherwise, the organism state of SubSeq was set to TaxAmbig. The task outlined in FIG. 3 was applied to each SubSeq() of ProSeq. In all cases, the identified organism was assigned to each SubSeq() representing the taxonomic class, which is the common parent of all constituents in Rank1.

After examining each SubSeq, the algorithm determined the organism to identify ProSeq from SubSeq, and moved to the next task (FIG. 4). When all the components of SubSeq have an identified organism value of Null, the ProSeq is negative and the next ProSeq is investigated. For ProSeq, if there is only a single component in SubSeq or all components in SubSeq have the same identified organism, a Result1 entry is formed for the identified organism, and the organism specificity is TaxUnique for multiple SubSeq entries, or a single SubSeq It inherits the state of the entry. If there were multiple entries in SubSeq with different identified organisms, further analysis was performed. SubSeq was then rearranged by MaxScore (bit score) so that the components with the highest two best scores were SubSeq(1) and SubSeq(2). When SubSeq(1) has a score of ≥ 30% (score ratio threshold) than that of SubSeq(2), the ProSeq inherits the organism-specific and identified organisms of SubSeq(1). Otherwise, the organism status of ProSeq was TaxAmbig and the identified organism was a taxonomic class, a common model of all subsequences. If all subsequences were contained within only two taxonomic classes, the direct subclass and the model class, the identified organism was an organism of subsequence within the subclass. The subtask included in FIG. 4 was terminated and the task I loop was continued. A list of ProSeq with detected organisms was formed in the Result1 array.

All pathogen identification and positive call

After task I was terminated, the values of the identified organisms listed in Result1 were examined using Task II (see Fig. 1), and when the values identified the same taxonomic class, they were grouped together. Each entry in Result1 was examined, and a new entry was formed in Result2 if the identified organism did not appear in the list above. In most cases, the entry in Result2 represents the individual organism detected, but could still contain redundant information. An entry in Result2, one with an identified organism, a taxonomic model in the other, could represent virtually the same pathogen. The same identification could not have occurred because the genomic targets did not hybridize well to all of the ProSeq for a variety of possible reasons. Alternatively, two different, but closely related organisms were both able to hybridize to the microarray.

Although it was difficult to correlate the results from separate ProSeqs, Task III handled the final investigations and decisions as currently implemented. As a result of implementing the previous tasks in detail, we did not take into account information about what ProSeq was trying to detect. This made it possible to recognize cases in which the lower cases are positive and negative as well as indeterminate cases. In the final task, the algorithm considered whether ProSeq identifies organisms designed to be detected. The apparent negative ProSeq and the uncertain ProSeq were considered negative for the target pathogen. ProSeq's grouping for this was based on the grouping already performed in Task II. The entry of Result2 was looped. Pathogens in the targeted table were investigated using the entry's ProSeq. Identification of the Result2 entry If the organism was identical to the taxonomic class of the target pathogen or was of the same type, the Pathogen() array was updated with an entry positive for the targeted pathogen. When the Pathogen() array was invalid for the pathogen, the pathogen level of the identified organism was the level of the Result2() entry. If the entry had already been placed on the pathogen, then further comparison was required. Result2() and Pathogen entries were compared. In the case of a direct phenotypic relationship, the pathogen's identified organism was a phenotypic taxonomic class. Otherwise, a common model taxonomic class was recorded as benign identifying organisms. In most cases where all ProSeqs for pathogens were well hybridized, a sophisticated level of identification was recorded. However, if more than one ProSeq did not hybridize well, the recorded positive target pathogen was identified only at the genus or species level. The results of all three tasks were recorded to enable manual re-examination. It should be noted that organisms identified in Task II that do not belong to the target pathogen were recorded as non-target positive carriers. Details of what was identified in this case required investigation of Task II level results.

Pathogen identification

A sample of Chlamydia pneumoniae having 10 to 1000 genomic copies (reference 21 method) was selected to describe how to perform pathogen detection and identification when multiple ProSeqs were targeted against the same pathogen ( 21). RPM v1 possesses three highly conserved ProSeqs selected from genes encoding the major outer membrane proteins VD2 and VD4, and the DNA dependent RNA polymerase (rpoB) gene. HybSeq from different samples only differed depending on how many unique base calls were as shown in Table 1 below. The percentage of ProSeq called is 80-100%, except in one case where the distinctive call indicating that the assay detection limit has been reached is a concentration of 10 with only 11% rpoB ProSeq producing above this concentration. Was varied. The determinations performed for SubSeq at the end of each task for the various samples are shown in Table 1 below. ProSeq from different cases produced the same number of SubSeq. These SubSeqs from different samples scored different bit scores for the same top ranked return from BLAST. In fact, VD2 and VD4 produced exactly the same results. The NCBI taxonomic database classified the returns into four distinct groups representing the C. pneumoniae taxonomic group and the three-shaped strain group. AE001652, AE002167, AE017159, and BA000008 showed the return of all ProSeq for each sample because they represented the database entries of the fully sequenced genome. One rpoB SubSeq produced the organism-specific, SeqUniqu. All other SubSeqs were TaxAmbig, as multiple returns were returned from different taxonomic classes. Since each of the VD2 and VD4 ProSeq has a single SubSeq, Task I assigned the state of SubSeq to ProSeq. In the case of rpoB ProSeq, the bit score of one SubSeq was sufficiently large, and the algorithm assigned the identification of that SubSeq to ProSeq. Task II of the algorithm grouped all three ProSeqs together because they all possess the same identifying organism and assign TaxAmbig. The results of Task III were positive for the target pathogen C. pneumoniae, and this determination was straightforward because all ProSeqs were consistent with each other and belonged to the same taxonomic class of target pathogens. Although the rpoB ProSeq was SeqUniqu, this was not the final decision for Task II because the SeqUniqu, ProSeq, was not a phenotypic group and the other ProSeq was TaxAmbig. The three recognized sub-strains were scored identically, and the sequence selected for ProSeq was very conservative, indicating that differences between these strains were not allowed.

Algorithm Determination for C. pneumoniae at several concentrations for SubSeq, Tasks I, II, and III Genome
copy ProSeq Distinctive call #SubSeq SubSeq organism identification and peculiarity,
Beat score Task I Task II task
III 1000 VD2 89% One (G1) C.pne, TA, 145 C.pne TA C.pne TA

Positive value C.pne

VD4 91% One (G1) C.pne, TA, 145 C.pne TA rpoB 80% 2 (G2) C.pne, SU, 307
(G3) C.pne, TA, 73 C.pne TA 100 VD2 100% One (G1) C.pne, TA, 164 C.pne TA C.pne TA



VD4 97% One (G1) C.pne, TA, 156 C.pne TA rpoB 80% 2 (G2) C.pne, SU, 343
(G3) C.pne, TA, 87 C.pne TA 100 VD2 83% One (G1) C.pne, TA, 136 C.pne TA C.pne TA



VD4 91% One (G1) C.pne, TA, 145 C.pne TA rpoB 84% 2 (G2)C.pne, SU, 318
(G3) C.pne, TA, 82 C.pne TA 10 VD2 100% One (G1) C.pne, TA, 164 C.pne TA C.pne TA



VD4 97% One (G1) C.pne, TA, 156 C.pne TA rpoB 90% 2 (G2)C.pne, SU, 340
(G3) C.pne, TA, 89 C.pne TA 10 VD2 100% One (G1) C.pne, TA, 164 C.pne TA C.pne TA



VD4 93% One (G1) C.pne, TA, 148 C.pne TA rpoB 11% 0 Null, Null Null, Null (G1) J138 (BA000008), AR39 (AE002167), Tw-183 (AE017159), Cpne
(M69230,AF131889,AY555078,M64064,AF131229,AF131230)
(G2) Cpne (S83995)
(G3) J138 (BA00008), AR39 (AE002167), Tw-183 (AE017159)
SU: Abbreviation for SeqUniqu
TA: Abbreviation for TaxAmbig

Influenza and human adenovirus (HAdV) were the only pathogens carrying the selected ProSeq to tolerate detailed strain level differences as discussed in previous studies (19, 20, 21). In this previous study using passive assays, it was found that the microarray results are in very good agreement with conventional sequencing results for clinical samples. The results of running the CIBSI 2.0 program using the updated NCBI database for the raw microarray results were compared with the previous results (Table 2). The organisms identified were not identical to the original results due to differences in the database used. In fact, the conventional sequencing results sent from this document to the NCBI were found for all samples present during the return with the best score. For 8 out of 13 influenza A and 3 out of 12 influenza B, the results of tasks I and II confirmed that the conventional sequencing was the single best return and thus was the identified organism. It was not surprising that, due to the large number of separate sequences in the database for the hemagglutinin gene, no single distinct entry was found in some cases. For each of the remaining five influenza A samples, the other sequence returned differed from the conventional sequence by less than 0.2%. Fewer samples using specific isolation identification for influenza B were due to the use of an older reference sequence for ProSeq, allowing less hybridization to occur (19). In addition, this meant that when multiple sequences were carried for the sample, they exhibited a greater genetic variation by up to 2%. This comparison only showed algorithmic analysis at the task I level for hemagglutinin (HA) ProSeq, which was a conventional sequenced region. As previous studies did not seek to obtain consensus from multiple ProSeqs, there could be no comparisons for Task III results. As a result of current methods of masking Task III level identification, organisms recorded at this level were less specific (H3N2 or Flu B) for all samples (Appendix Tables 1A and 1B). For the HAdV sample, the algorithm also reproduced (not shown) finer scale differences that were previously performed by the passive method.

Pathogens Identified from HA ProSeqs of Influenza A and B Clinical Samples Sample name By prior literature
Strain identification GenBank
Registration Number Identification of HA3 ProSeq by CIBSI 2.0 GenBank
Registration Number A/Colorado/360/05 A/Nepal/1679/2004 (H3N2) AY945284 A/Colorado/3
60/05 ^* DQ265717 A/ Qatar/2039/05 A/Nepal/1727/2004 (H3N2) AY945272 A/Qatar/203
9/05 DQ265707 A/Guam/362/05 A/Nepal/1679/2004 (H3N2) AY945264 A/Guam/362/05 DQ265715 A/Italy/384/05 A/Nepal/1727/2004 (H3N2) AY945272 A/Italy/3
84/05 DQ265713 A/Turkey/2108/05 A/Nepal/1664/2004 (H3N2) AY945265 A/Turkey/2108/
05 DQ265718 A/Korea/298/05 A/Nepal/1727/2004 (H3N2) AY945273 A/Korea/298/0
5 DQ265710 A/Japan/1337/05 A/Malaysia/2256/2004 (H3N2) ISDN110616 A/ Japan /1337/
05 ^* DQ265712 A/Japan/1383/05 A/Malaysia/2256/2004 (H3N2) ISDN110616 A/Japan/1383/
05 DQ265711 A/Ecuador/1968/04 A/New York/17/2003 (H3N2) CY001053 A/Ecuador/1
968/04 ^* DQ265716 A/Iraq/34/05 A/Christchurch/178/2004 (H3N2) ISDN110530 A/Iraq/34/
05 DQ265714 A/Peru/166/05 A/Macau/103/2004 (H3N2) ISDN64772 A/Peru/166/0
5 DQ265708 A/New York/2782/04 A/New York/391/2005 (H3N2) CY002056 A/New York/2782/
04 ^S* DQ265709 A/UK/400/05 A/New York/227/2003 (H1N1) CY002536 A/UK/2005 ^* DQ265706 (continue) (continue) Sample name By prior literature
Strain identification GenBank
Registration Number To CIBSI 2.0
By HA3
ProSeq identification GenBank
Registration Number B/Peru/1324/
04 B/Milan/66/04 AJ842082 B/Peru/1324/
04S* DQ265728 B/Peru/1364/
04 B/Milan/66/04 AJ842082 B/Peru/1364/
04S* DQ265726 B/Colorado/2
597/04 B/Texas/3/2002 AY139049 B/Colorado/2
597/04S* DQ265724 B/Japan/1905/
05 B/Texas/3/2002 AY139049 B/Japan/1905/
05S* DQ265727 B/Japan/1224/
05 B/Texas/3/2002 AY139049 B/Japan/1224/
05S* DQ265719 B/Alaska/1
777/05 B/Texas/3/2002 AY139049 B/Alaska/1
777/05S DQ265730 B/UK/1716/
05 B/Texas/3/2002 AY139049 B/UK/1716/
05S DQ265723 B/UK/2054/
05 B/Texas/3/2002 AY139049 B/UK/2054/
05* DQ265722 B/Hawaii/199
0/04 B/Tehran/80/02 AJ784042 B/Hawaii/199
0/04 DQ265721 B/Hawaii/199
3/04 B/Tehran/80/02 AJ784042 B/Hawaii/199
3/04S* DQ265720 B/Arizona/1
48/04 B/Tehran/80/02 AJ784042 B/Arizona/1
48/04* DQ265725 B/Arizona/1
46/04 B/Tehran/80/02 AJ784042 B/Arizona/1
46/04* DQ265729 ^* Multiple returns are fixed with these returns.
^S HybSeq is split into multiple SubSeqs.

The following detection example for the Mycoplasma pneumoniae pathogen describes the case where there was only a single ProSeq for the target pathogen, which is the result of task II automatically identifying the organism for task I of the algorithm and targeting this. It was meant that it was the only ProSeq considered in Task III for the pathogen to be treated. This ProSeq is also not suitable for microscopic identification because it was selected from the highly conserved region (345 bp) of the cytadhesin P1 gene. Forty microarrays were tested with the same purified nucleic acid stock and in all cases recognizing M. pneumoniae or its substrain taxonomic database entries were fixed in MaxScore. To further understand this return, the database sequence was examined, and the sequence was partially split into 3 groups A, B, and C based on how well the database sequence matched the reference sequence used to generate the ProSeq. I did. The placement of the database entries in the 3 groups was determined from the CLUSTAL alignment of this gene sequence. This alignment confirmed that the database entries differed more significantly from each other in the region where the database entries were not represented by ProSeq and contained sufficient variability to allow for finer differences. The members of group A matched exactly with ProSeq and were indistinguishable from each other on the microarray. Similarly, members of group B were consistent with ProSeq except for the 199th position where the called base was C rather than T. The Group C sequence was more variable and contained some database entries that could be distinguished from other entries in ProSeq. In the case of 40 experimental tests with M. pneumoniae in which 95% of ProSeq was hybridized, only 65% of the results had a clear base call at the 199th position. When the base call is clear, it always coincided with the group B sequence. When an N base call was generated at position 199, both group A and B sequences were returned with the same score. Regardless, the target pathogen identified as a positive value was M. pneumoniae for all samples tested.

These examples showed how the decision is made irrespective of whether a single or multiple ProSeq is applied to the target pathogen. These examples also demonstrated that the possible level difference is strongly determined by the quality of the chosen ProSeq. For some pathogens, subtle level differences may not be required, and selections currently tested on RPM v1 will provide satisfactory information. The CIBSI 2.0 algorithm has proven its ability to automatically record the maximum level difference that can be supported by HybSeq information.

Genetic Near Neighbors

To illustrate how the algorithm handles closely related genetic species, a sample of untargeted pathogens was considered. In the case of soybean virus, which is one of the bio-threatening pathogens, on RPM v1, the feasibility study proved that the detection of soybean virus DNA template was always identified as a positive value. The array carries two ProSeqs derived from cytokine response modifier B (VMVcrmB, ˜300 bp) and hemagglutinin (VMVHA, ˜500 bp) for soybean virus detection. Table 3 below shows the results of 18 runs of each ProSeq in which the closely adjacent vaccinia virus was spiked by nasal washing at various concentrations. The percentage of ProSeq hybridized is sufficient in that it can be inferred that these tiles identify the presence of the target if only the hybridization pattern is considered. This indicated that the selected reference sequence was not the best choice. However, when the algorithm was applied, none of the samples were in fact identified as soybean or smallpox virus. Vaccinia virus has always been one of the described orthopoxvirus species with the highest score for VMVcrmB ProSeq, but it was the only one that was detected as a possible species in only 7 cases. In three samples with the lowest concentration and fragment of VMVcrmB hybridization, this ProSeq identified soybean virus as one of a number of orthopoxvirus species that may be responsible for hybridization. The lower limit of detection for the amplification method used was between this concentration and above. The VMVHA ProSeq identified orthopoxvirus species in only two experiments and described soybean virus as one of the returns with the fixed best score. In both cases, VMVcrmB ProSeq specifically identified vaccinia virus as the best match. The percentage of hybridized ProSeq correlated with the concentration of the sample.

Identification of organisms from vaccinia samples on soybean virus ProSeq CFU ProSeq VMVCRMB VMVHA % Identity % Identity 5*10 ⁷ 77.9 Vaccinia virus 29.4 Orthopox virus 5*10 ⁷ 79.8 Vaccinia virus 25.7 Orthopox virus 1.6*10 ⁷ 79.4 Vaccinia virus 14.8 - 1.6*10 ⁷ 77.5 Orthopox virus ^* 24.5 - 1.6*10 ⁷ 76.8 Vaccinia virus 21.6 - 1.6*10 ⁷ 74.5 Orthopox virus ^* 17.3 - 5*10 ⁶ 77.9 Vaccinia virus 25.7 - 5*10 ⁶ 78.3 Orthopox virus ^* 22.0 - 5*10 ⁶ 73.0 Vaccinia virus 13.0 - 5*10 ⁶ 73.4 Orthopox virus ^* 7.8 - 1.6*10 ⁶ 75.3 Orthopox virus ^* 8.6 - 1.6*10 ⁶ 49.8 Vaccinia virus 6.6 - 1.6*10 ⁶ 65.5 Orthopox virus ^* 10.0 - 1.6*10 ⁶ 62.9 Orthopox virus ^* 8.2 - 5*10 ⁵ 58.4 Orthopox virus ^* 9.0 - 5*10 ⁵ 56.2 Orthopox virus 8.0 - 5*10 ⁵ 49.0 Orthopox virus 9.3 - 5*10 ⁵ 44.6 Orthopox virus 7.8 - ^* -CFU of only neighboring species within orthopox virus, not soybean or smallpox virus-Colony forming unit

Filtering

The above example explained the importance of the filtering part of the algorithm, taking into account the H1N1 neuraminidase (NA1) from the human influenza A/Puerto Rico/8/34 (H1N1) strain and the HybSeq of ProSeq for the matrix gene. The reason for filtering is that ProSeq's HybSeq is transferred to BLAST of a single query, especially when using the BLAST parameter that maximizes the use of base calls, biasing scores for strains with insertions or deletions relative to ProSeq. Because you can. The sliding window test was part of the algorithm that controlled filtering. When filtering was stopped, the entire HybSeq was used for the single partial sequence for the two influenza ProSeq showing significant hybridization. The A/Weiss/43 (H1N1) strain was identified as the most possible strain derived from HybSeq of NA1 ProSeq, while HybSeq of matrix ProSeq accurately identified A/Puerto Rico/8/34. In order to further understand the source of biasing, the reference sequence used to generate ProSeq and the CLUSTAL alignment of the two strains of NA1 genes are shown in FIG. 5. The two strains showed 95% identity (67 mismatch among 1362 bases aligned), but compared to A/Puerto Rico/8/34 (SEQ ID NO: 3), A/Weiss/43 (SEQ ID NO: 2) And NA1 ProSeq (SEQ ID NO: 1) There was a 45 base stretch inserted into both. Filtering the defaults, the NA1 ProSeq was split into 5 SubSeqs, as the algorithm encountered a large stretch without any calls. In task I, the algorithm is based on the identification organism of H1N1 as several isolates, including the A/Puerto Rico/8/34 strain, with three shorter SubSeqs fixed to the best score, while the other two SubSeqs are the most. It was determined to have an identified organism of only closely matched A/Puerto Rico/8/34 strains. The organism identified by NA1 ProSeq was A/Puerto Rico/8/34 because one of the SubSeq had a much higher score. The ProSeq supported the same strain identification done in the matrix ProSeq. The organisms identified were A/Puerto Rico/8/34 because two ProSeqs detected only these organisms. The exact target pathogen was detected by filtering, whereas in the absence of filtering, the level of identification of the target pathogen was influenza A (H1N1 subtype), because of the two organisms, A/Puerto Rico/8/34 and A/Weiss/43. This was because it was detected. Separation of HybSeq into SubSeq to eliminate biasing can reduce the level of identification, as occurs for 3 out of 5 SubSeqs in this case. A preceding example for vaccinia was another example where the identification of the wrong species (Camel Pox or Callithrix jacchus) could occur when filtering was not used. The clinical samples in Table 2 are a number of SubSeqs. It was shown that the HybSeq separated by is capable of very specific identification.

As a secondary point, when using various strategies for amplification different from the general one, it was necessary to perform additional filtering as described in this method to remove potential biasing from specific primers. Figure 5 has the original (SEQ ID NO: 4) results and shield filtered (SEQ ID NO: 5) results for hybridization of A/Puerto Rico/8/34 to show an example of such interference. In addition to the problem with the bias due to the above-described reason, there is a sequence of 18 bases present in the original result in which N is formed after filtering because it is at a position where it interacts with the primer. If these base calls are included in the constructed subsequence, queries to ProSeq will favor the incorrect strain.

The algorithm successfully provided pathogen identification for the maximum level of detailed possible (species or strains) depending on the quality of each ProSeq. This identification performance requires minimal input for the identity of the pathogen, which is possible for non-specialist use. An important feature is the use of a taxonomic database that categorizes organisms into ordered groups and provides relationships between organism entries, allowing for the elimination of redundancy, comparison of different related prototypical sequences, and simplification of data presentation. It was specified to be fully automated. This makes it possible for the database, ie NCBI, to be redundant and subject to minimal curation, but continue to accommodate new sequences that have been updated to be used most successfully. This has been described as using only the NCBI database, but other databases or ordinary ones can be easily used, and performance can be improved. The algorithm could provide accurate identification at all levels of analysis for pathogens represented by less variable or highly conserved ProSeq. For more variable or rapidly mutating pathogens, such as influenza A virus, tasks I and II, it still provides accurate detailed identification, but task III was unable to record fine scale differences. Comparison of conventional sequenced influenza virus gene sequences demonstrated that the algorithm can be automatically adjusted for updates in the database. The algorithm has demonstrated its ability to properly differentiate hybridization on ProSeq caused by a specific pathogen from what is caused by a genetically close (adjacent) strain, and is inaccurate while eliminating one potential cause of false positives. I didn't sympathize. Filtering the original hybridization results served to reduce the computational time that accounts for potential primer interference, and more importantly, reduced potential biasing. This simple integration algorithm provides sufficient and accurate identification for immediate use of RPM v1 or similar rearrangement arrays and analyzes.

In addition to describing the success of the CIBSI 2.0 program, the work involved in developing the algorithm provided insight into the importance of selecting an appropriate ProSeq. RPM v1 is a first rearrangement array specifically designed for detection of multiple pathogens using database similarity search, and was provided as a prototype for the present application. It was demonstrated that, if designed correctly, a single ProSeq as small as 100 bp may be sufficient to clearly identify organisms. However, it is clearly indicated that several longer ProSeqs provide better confirmation and more detailed information of the pathogen. The design emphasis in this regard was generally on the performance applicable to any pathogen. Improvements to the performance of Task III may require more information on individual pathogens and may need to be developed for each specific pathogen or class of pathogens. In addition, this information may require algorithms to identify that differences between sample and database entries indicate significant mutations. The hierarchical design of data analysis can facilitate incorporating analyzes that are formed according to analyzes that have already been performed. The use of properly designed rearrangement microarrays and this automated detection algorithm allows testing of multiple organisms simultaneously while providing micro-strain level differences while accessing detailed strain recognition, antibiotic resistance markers, and information on pathogenicity. It can provide directions for developing an analysis that can be done. This includes differential diagnosis of diseases with multiple potential causes (i.e. febrile respiratory diseases), tracking of emerging pathogens, discrimination of biological threats from innocuous genetic neighbors in the field of surveillance, and tracking the effects of co-infection or multiple infections. It will enable the analysis of partial sequence information from multiple organisms for the same field. The concept of recording and categorizing different degrees of identification depending on the target sequence set and the quality of the sample is not limited to rearrangement microarrays, and any sequence level call capable of broadcasting a sequence level call that can be used to query a reference DNA database. It is generally more applicable to the platform. As the trend toward analysis testing for multiple pathogens increases, automated analysis tools, such as such, have become more important for rapid identification in a simple format useful for non-experts who are dealing with it on the day-to-day basis.

Source code

Below is a listing of the PERL source code of an embodiment of the disclosed method. The "overclinical" program is a top-notch program that runs other programs. "fstorepi" performs filtering, sequence generation, and query file generation. This program is used for the input file "primerhyb.dat" which contains the position of a predetermined list that is changed to N. "runblast" executes a BLAST query. "dbparse" performs taxonomic analysis. The program uses an input file "chip1pathogengroups" containing a list of target pathogens for each ProSeq.

Obviously, many modifications and variations of the present invention are possible in light of the above teaching. Therefore, it will be appreciated that the claimed invention may be practiced in addition to as specifically described. Any criteria claiming components that use the singular, for example, the items “a,” “an,” “the,” or “said” are not to be construed as limiting to the singular component.

SEQUENCE LISTING <110> Malanoski, Anthony P Lin, Baochuan Schnur, Joel M Stenger, David A <120> COMPUTER-IMPLEMENTED BIOLOGICAL SEQUENCE IDENTIFIER SYSTEM AND METHOD <130> 97748US2 <150> 60/691,768 <151> 2005-06-16 <150> 60/735,876 <151> 2005-11-14 <150> 60/735,824 <151> 2005-11-14 <150> 60/743,977 <151> 2006-03-30 <150> 11/177,647 <151> 2005-07-02 <150> 11/177,646 <151> 2005-07-02 <150> 11/268,373 <151> 2005-11-07 <150> 11/422,425 <151> 2006-06-06 <150> 11/422,431 <151> 2006-06-06 <160> 5 <170> PatentIn version 3.3 <210> 1 <211> 61 <212> DNA <213> Human Influenza A <220> <221> gene <222> (1)..(61) <223> NA1 <400> 1 ctgggtaaat caaacatatg tcaatattaa caacactaac gttgttgctg gaaaggacac 60 a 61 <210> 2 <211> 61 <212> DNA <213> Human Influenza A <220> <221> gene <222> (1)..(61) <223> NA1 <400> 2 ctgggtaaat caaacatatg ttaatattag caacactaac gttgttgctg gaaaaggcac 60 a 61 <210> 3 <211> 16 <212> DNA <213> Human Influenza A <220> <221> gene <222> (1)..(16) <223> NA1 <400> 3 ctgggtaaag gacaca 16 <210> 4 <211> 61 <212> DNA <213> Unknown <220> <223> Raw data <220> <221> misc_feature <222> (1)..(61) <223> n is a, c, g, or t <400> 4 ctgggnnnnn nnnnnnnnnn nnnnnnnnnn nnnnnnngnc gttgttgctg gaaaggncac 60 a 61 <210> 5 <211> 61 <212> DNA <213> Unknown <220> <223> Filtered data <220> <221> misc_feature <222> (1)..(61) <223> n is a, c, g, or t <400> 5 ctgggnnnnn nnnnnnnnnn nnnnnnnnnn nnnnnnnnnn nnnnnnnnnn nnnnnnnnac 60 a 61

Claims (9)

A method of processing a biological sequence obtained from an analysis,
Masking base calls that are located in a predetermined list of positions in the biological sequence as ambiguous (N) base calls, wherein the unique base calls in the biological sequence are not ambiguous (N) base calls, The entry indicates the ability of the substance to hybridize to the microarray used to generate the biological sequence, wherein the substance is not a nucleic acid of the target pathogen; And
Determining the ratio of single nucleotide polymorphisms in the biological sequence relative to the reference sequence to the total number of unique base calls
&Lt; / RTI &gt; The method of claim 1, wherein the substance is a PCR primer. The method of claim 1, further comprising: selecting a potential partial sequence of initial length from the biological sequence when the ratio of the single nucleotide polymorphism is below the SNP (single nucleotide polymorphism) threshold; And
Calculating the total content of the unique base call in the potential partial sequence
&Lt; / RTI &gt; The method of claim 3, wherein the SNP threshold is about 20%. The method of claim 3,
Extending the potential partial sequence by one base if the total content of the unique base call is above the unique base call threshold;
Recalculating the total content of unique base calls in the extended potential partial sequence;
Repeating extension and recalculation until the total content of the unique base call is less than the unique base call threshold;
Removing an ambiguous (N) base call that is trailing from the extended potential partial sequence to form a query partial sequence; And
Calculating the total content of the unique base call in the query subsequence
&Lt; / RTI &gt; 6. The method of claim 5, wherein the unique base call threshold is about 40%. 6. The method of claim 5, wherein repeating the extending and recalculating step stops if the last 21 positions of the extended potential partial sequence have less than 4 unique base calls. The method according to claim 5, wherein if the length of the query subsequence and the total content of the unique base call in the query subsequence meets a predetermined requirement, the query subsequence is transferred to the genetic database as a query to obtain a biological sequence. Further comprising the step of identifying. The method of claim 8, wherein the requirement is
The query subsequence comprises at least seven consecutive unique base calls; And
The length of the query subsequence is greater than 100 bases, or the length of the query subsequences is 30 to 100 bases, and the total content of unique base calls in the query subsequences is at least about ((length-30 of the query subsequences) * 0.2857 + 70)% or the query subsequence is less than 30 bases in length and the total content of unique base calls in the query subsequence is at least about 95%
How to include.

Images