Method for optimized compression of quality markers, and computer device having a compression unit

The method optimizes genome sequence quality marker compression by evaluating trust values and quantizing based on them, using entropy coding and smoothing, effectively reducing data while minimizing information loss, addressing the challenge of direct correlation between compression and information loss in existing methods.

WO2024231470A1PCT designated stage expired Publication Date: 2024-11-14GOTTFRIED WILHELM LEIBNIZ UNIV HANNOVER
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
PCT/EP2024/062773
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-11
Filing Date
2024-05-08
Publication Date
2024-11-14

AI Technical Summary

Technical Problem

Existing methods for compressing quality markers in genome sequences often result in a direct correlation between compression degree and information loss, leading to a loss of quality during further processing.

Method used

A method involving a compression unit that reads quality markers, applies an evaluation model to determine a trust value, and quantizes them based on this value for optimized compression, using entropy coding and potentially smoothing the results, to minimize data quantity while maintaining acceptable information integrity.

Benefits of technology

This approach achieves the smallest possible data reduction with minimal information loss, particularly effective since quality markers can comprise up to 80% of genome sequence data, offering a clear advantage in data reduction with adapted compression.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure EP2024062773_14112024_PF_FP_ABST
    Figure EP2024062773_14112024_PF_FP_ABST
Patent Text Reader

Abstract

The invention relates to a method using a compression unit for optimized compression of quality markers of a genome sequence determined from a genome, the genome sequence having respective genome information at respective genome positions and respective quality markers being assigned to the respective genome position and having a quality value for genome information at the respective genome position in the genome sequence. Moreover, the invention relates to a computer device having a compression unit.
Need to check novelty before this filing date? Find Prior Art

Description

Method for optimized compression of quality markers and computer equipment with a compression unit

[0001] The invention relates to a method for the optimized compression of quality markers of a genome sequence determined from a genome by means of a compression unit, wherein the genome sequence has respective genome information at respective genome positions and the respective quality markers are assigned to the respective genome position and have a quality value of genome information at the respective genome position in the genome sequence. Furthermore, the invention relates to a computer device with a compression unit.

[0002] Methods exist for compressing quality markers of a genome sequence determined from a genome. However, with these methods, there is a direct correlation between the degree of compression and the loss of information during further processing of the genome sequence.

[0003] For example, so-called heterogeneous consensus algorithms are used to generate consensus genome information for every genome position. Furthermore, compression methods are known in which quality information, such as quality markers, is nested or smoothed using a smoothing function. However, a loss of quality often becomes apparent, especially during the further processing of such quality markers. Furthermore, Methods are known in which quality markers are modeled using a first-order Markov process.

[0004] US 10,938,414 B2 describes a method for encoding and decoding quality markers, using an encryption table.

[0005] US 10,691,775 B2 describes a so-called bioinformation system, apparatus, and methods for analyzing genetic information.

[0006] Documents CN 111640467 A and CN 10539145 A also concern data compression.

[0007] WO 2020 / 191390 A2 reveals the allocation of quality markers to genome sequences.

[0008] WO 20157180203 Al describes a lossless compression system for quality markers and a corresponding procedure.

[0009] WO 2018 / 068845 Al and WO 2019 / 012153 Al each disclose methods for encoding and decoding quality values ​​of a data structure.

[0010] US 2021 / 295950 Al reveals another method for compressing quality markers.

[0011] James K. Bonfield ET AL: "Compression of FASTQ and SAM Format Sequencing Data", PLoS ONE, Vol. 8, No. 3, January 1, 2013 (2013-01-01), page e59190, XP055330942, also discloses a method for compressing information.

[0012] Furthermore, IDOIA OCHOA ET AL: "QualComp: a new lossy compressor for quality scores based on rate distortion theory", BMC BIOINFORMATICS, BIOMED CENTRAL , LONDON, GB, Vol. 14, No. 1, June 8, 2013 (2013-06-08) , page 187, XP021154440, reveals a method for compressing information.

[0013] The purpose of the invention is to improve the state of the art.

[0014] The task is solved by a method for the optimized compression of quality markers of a genome sequence determined from a genome using a compression unit, wherein the genome sequence contains respective genome information at respective genome positions and respective quality markers are assigned to the respective genome position and have a quality value of genome information at the respective genome position in the genome sequence, with the following steps: - Reading the respective quality marker for the respective genome position into the Compression unit, so that the respective quality marker is present in the compression unit at the respective genome position, - Applying an evaluation model to determine a confidence level for the respective quality marker at the respective genome position, such that a confidence level of the quality marker at the respective genome position exists, - Quantizing the respective quality marker for the respective genome position based on the Trustworthy, whereby a The quantization intensity is chosen depending on the confidence level, so that a quality marker of the respective genome position is quantized based on the confidence level. - Compressing the quantized quality marker so that the quality markers are optimally compressed.

[0015] A key concept of the invention is to first perform an evaluation to determine a confidence level for each quality marker, and then, depending on this confidence level, to select a targeted and individualized quantization. Overall, it has been shown that this approach achieves the smallest possible data volume with acceptable information loss. It should be noted that, in particular, the quality markers for a given genome position can account for up to 80% of the data from a genome sequence, so that a tailored and individually selected compression can offer a significant advantage in reducing the data volume.

[0016] The following terms are explained in this context:

[0017] A "method for optimized compression of quality markers" serves to improve the compression of quality markers compared to known methods. It deals with... This refers to an instruction that can be executed, for example, by computers to perform "compression." Such compression describes the reduction of a total data volume compared to an initial data volume. Ideally, compression allows for no loss of information, but depending on the compression algorithm, some loss of information must be accepted in order to compress a data volume. The goal of compression is always to minimize the reduction of the data volume while keeping information loss to a minimum. The quality of the compression can be evaluated, depending on the application, in terms of the highest possible data reduction, the most lossless compression possible, or a compromise between these objectives.

[0018] "Quality markers" are individual pieces of information that serve, in particular, to establish and / or maintain the reliability of information at a specific position within a dataset. In this context, a "genome" is understood as a coherent set of genetic information, for example, from a living organism. The genome is usually "sequenced," whereby corresponding subsets of the genome are determined as a "genome sequence." Such genome sequences are also referred to as "reads" in technical terminology, whereby corresponding genome sequences are determined starting from different positions within the genome and, for example, overlapping positions in the These positions can refer to a genome sequence. Such positions, called "genome positions", denote a respective address within a genome sequence and / or within the genome, whereby the genome in this context is to be understood as a chain of genome information provided with individual addresses.

[0019] "Genetic information" refers to an identifier of a particular genome in the form of a designation of nucleobases, which are commonly designated, for example, by the letters A, C, G, and T, thus representing the different nucleobases adenine, cytosine, guanine, and thymine of DNA. Such genome information can be extracted from either DNA or RNA.

[0020] It should be mentioned that such a genome sequence can be determined using laboratory equipment, but may be prone to errors due to the design of the equipment. For example, a single nucleobase may be incorrectly read, resulting in an error in the generated sequence of nucleobases at the corresponding position in the "read." Consequently, modern instruments for determining the genome sequence provide a quality marker for each genome position in a given read, indicating the quality of the corresponding value for the genome information within the sequence. For instance, unreliable genome information is indicated by a different quality marker than highly reliable genome information. This is particularly important for the These quality markers are of high importance for the further processing of corresponding genome sequences.

[0021] It should also be mentioned that corresponding genome sequences, provided they represent overlapping positions within the genome, can be arranged in parallel and offset from each other, thus enabling a comparison between different genome information data from different genome sequences for the respective common genome position. The same applies to associated quality scores or quality markers.

[0022] The "quality score" of a piece of genomic information is, for example, the numerical value of the respective quality marker, which can be expressed in any format. For instance, such a quality score can be a percentage between 0% and 100%, but it can also be represented differently, such as using a letter code or a numerical value on a freely chosen scale. The core message of the quality score is the reliability of the respective piece of genomic information.

[0023] "Reading" the respective quality marker for the respective genome position refers to transferring data into a compression unit, where the compression unit is implemented, for example, in a computer and executed using an algorithm. However, such a compression unit could also be, for example, a mechanical switching arrangement. to process data information, even if this case is unlikely in physical implementation.

[0024] Applying an evaluation model describes the process of applying an evaluation model, i.e., an algorithm for determining the confidence level of a given quality marker. This can involve, for example, comparing different quality markers at different genomic positions. Furthermore, different genomic information from different genomic positions can also be included in the comparison. The resulting confidence level serves as a measure of the reliability of the quality value assigned to the quality marker. For instance, the evaluation model might indicate that the confidence level for a given quality marker is very low, meaning that the quality value assigned to that marker is not considered reliable.

[0025] According to the invention, the respective quality marker is then "quantized" based on the confidence level. For example, if the confidence level of the respective quality marker is high, a comparatively fine quantization is chosen, where the strength of the quantization is referred to here as "quantization intensity," with a high quantization intensity being equivalent to a coarse quantization and vice versa. If the corresponding confidence level is very low, a low quantization intensity may be used. This is chosen to avoid further degrading unreliable quality information through quantization. Similarly, depending on the application, an inverse application is also conceivable, in which, for example, a lower quantization intensity is chosen for quality markers with a high confidence value, for instance, to avoid distorting a high information content through compression.

[0026] Subsequent "compression" then refers to reducing the amount of data containing the quality markers, for example to store the data with lower storage requirements.

[0027] The result after compressing the quantized quality markers is "compressed quality markers," meaning data information regarding the quality markers with a smaller data volume. It should be noted that different data loss rates can be accepted during quantization and compression, depending on the confidence level.

[0028] To reliably determine a confidence level, the evaluation model is a genotype-match model, in which, in particular, an uncertainty profile is determined for a sequence of genome positions. Such an uncertainty profile reveals corresponding uncertainty values, i.e., a corresponding inverse statement about the confidence level.

[0029] Alternatively and / or additionally, the assessment model, according to one implementation form, is an activity-based one. A tracked model is used, in particular to determine an activity profile for a sequence of genome positions. For this purpose, a frequency distribution can be used as a simple model, as well as any further, more complex evaluation.

[0030] A "genotype match model" determines the probability distribution of all theoretically possible genotypes for each genome position. A statistical model is used for this purpose. The result is an uncertainty profile that determines the uncertainty of the respective genomic information for each genome position. This indicates the probability that a given piece of genomic information is correct. Depending on this probability, the quantization intensity of the quality marker for a given genome position can then be chosen. The uncertainty of a genome position can be considered a metric that measures the probability that a particular genotype is indeed the biologically actual genotype. Therefore, the uncertainty can be used to determine the acceptable level of data loss for all quality markers at a genome position.

[0031] An “activity-based tracked model”, which can be used as an alternative to the aforementioned genotype-match model, uses systematic assumptions to determine an activity profile across the respective positions. These systematic assumptions include an improved model for estimating the The probability that some or more of the respective nucleobases are defective. Furthermore, it is a systematic assumption that genome sequences where some genetic information cannot be assigned to genome positions contribute to increased uncertainties in their genome position environment.

[0032] In particular, the confidence level of a given quality marker can be determined by assigning it to a specific confidence range, especially with weighting. For example, a set of possible confidence levels, such as between 0% and 100%, is divided into corresponding intervals or sections, with each confidence level then assigned to a specific interval segment, also with weighting. Thus, for example, quality markers in the range of high confidence levels, such as between 100% and 65%, can be treated similarly, whereas at significantly lower confidence levels, such as below 65%, a finer nesting is used to better capture corresponding deviations. The percentage values ​​given here are examples and may vary depending on the application.

[0033] According to one implementation method, compression is performed using entropy coding. For example, compression can also be achieved by further quantizing the quality markers.

[0034] Entropy coding is a method for lossless data compression, in which a string consisting of individual characters is converted into a bit sequence. For example, Huffman coding or arithmetic coding is used for this purpose. Alternatively, a sequence of symbols from a symbol set can be used instead of a bit sequence.

[0035] According to one implementation method, determining the confidence level involves comparing several genome sequences. This comparison is achieved by aligning the respective genome positions such that the corresponding genome positions in each sequence represent the same positions within the genome sequence. Thus, genome sequences can be arranged in such a way that corresponding genome positions in different genome sequences are, in effect, parallelized. For example, a comparison can be made between corresponding genome information at respective genome positions until the highest possible degree of agreement is achieved. The genome sequences are essentially "overlaid" in their positions, allowing for a direct comparison of the genome information and quality markers, particularly at each position.In particular, the respective quality markers can then be compared with each other, so that optimized compression can take place.

[0036] To further reduce the amount of data in the compressed quality markers, the following steps are used to store the data: compressed quality markers are offset, wherein the respective compressed quality markers within a character table are shifted by means of the offset in such a way that a single-digit representation is enabled, in particular in an ASCII character set.

[0037] This implementation form serves in particular to avoid, for example, multi-digit ASCII characters for representing quality markers and to place all possible quality markers in an area of ​​the ASCII character set that can be represented with a single digit.

[0038] According to one implementation method, further smoothing of the compressed quality markers and / or smoothing of the uncertainty profile, in particular by means of a Gaussian filter, can be carried out.

[0039] In another aspect, the problem is solved by a computer device with a compression unit which is configured to carry out a method according to one of the preceding claims.

[0040] Such a "computer device" is, in particular, a device for electronic data processing which is configured, for example, by means of an algorithm, to carry out the method according to the invention in accordance with one or more of the embodiments described above. It may, for example, be a personal computer, a data center, a server, or a comparable information technology device.

[0041] The invention will now be explained in more detail using exemplary embodiments. These will show... Figure 1 is a data matrix to represent Genome sequences with corresponding quality markers and derived confidence values.

[0042] A data matrix 101 shows, as an example, four different genome sequences, which are arranged in a matrix-like arrangement as genome sequences 102, 104, 106, and 108. Each genome sequence has a genome position 121, which, numbered sequentially, represents the position of a respective piece of genome information 103 in the genome underlying the genome sequence. At the same genome position 121, each piece of genome information 103 has a quality marker 105. The quality marker 105 is the result of an analysis of the genome sequence during its sequencing and provides information about the quality of the corresponding genome information. The quality marker also indicates the reliability of the information as a numerical value, represented here as an ASCII character.For example, a high quality marker indicates a high reliability of the read genome information 105 at the respective genome position 121.

[0043] The genome sequences 102, 104, 106 and 108 result as four so-called different "reads", i.e. a multiple successive sequencing of the same genome.

[0044] It should be noted that such a genome sequence cannot be read without errors; for example, at genome position 141, different reads at the same position can show different values ​​(here "G", "A"). Furthermore, the quality marker 105j also varies. Therefore, there is an overall uncertainty regarding the actual genome information present at position 141 in the underlying genome.

[0045] In particular, the quality markers 105 have a large amount of data, so their compression is desirable.

[0046] An analysis of the quality markers 105 is now performed to determine a confidence level. For this purpose, the quality marker is subjected to an evaluation model, in the example shown a genotype match model, so that confidence levels 131 are generated for the respective positions from this evaluation model. Different confidence levels indicate the reliability of the respective quality markers 105 across a number of genome sequences.

[0047] A “genotype match model” determines, as an example, the probability distribution of all theoretically possible genotypes for genome position 141. Assuming that the underlying genome is human and therefore diploid, this results in 10 possible genotypes, which By extracting the DNA sequence twice, without regard to order, the following results are obtained from the set of possible nucleobases: AA, AC, AG, AT, CC, CG, CT, GG, GT, TT. Considering the genomic information (G, A, A) present at genome position 141 with the associated quality markers ("&", "B", ">"), which express a quality rating in ascending order according to "poor", "moderately good", "moderately poor", a probability distribution is obtained in which several genotypes appear probable, e.g., AA and AG, but also AC and AT, due to the "poor" quality rating of the genomic information G. As a result, the uncertainty at genome position 141 is comparatively high.

[0048] The quality markers 105 are then quantized based on the determined confidence values ​​for each respective genome position 121. The result is quantized quality markers 151, which (see Figure 1) exhibit significantly higher uniformity than the original quality markers 105 in the data matrix 101.

[0049] The quantized quality markers are then combined into a compression, for example entropy coding.

[0050] The result is then compressed weighted Quality markers are present. Reference character list 101 Data Matrix 102 Genome sequence 103 genome information 104 genome sequence 105 quality markers 106 Genome sequence 108 Genome sequence 121 Genome position 131 Confidence level 141 Genome position 151 weighted quality markers

Claims

Patent claims:

1. Method for the optimized compression of quality markers (105) of a genome sequence (102, 104, 106, 108) determined from a genome by means of a compression unit, wherein the genome sequence (102, 104, 106, 108) is Genome positions (121) have respective genome information (103) and respective quality markers (105) are assigned to the respective genome position (121) and have a quality value of genome information (103) at the respective genome position (121) in the genome sequence (102), with the following steps: - Reading the respective quality marker (105) for the respective genome position (121, 141) into the compression unit, so that the respective quality marker (105) for the respective genome position (121, 141) is present in the compression unit, - applying an evaluation model to determine a confidence value for the respective quality marker (105) for the respective genome position (121, 141), so that a confidence value of the quality marker (105) for the respective genome position (121, 141) is available, - Quantify the respective quality marker to the respective genome position (121, 141) using the confidence value, wherein a quantization intensity is selected depending on the confidence value, so that a respective quality marker (105) of the respective genome position (121, 141) quantized based on the confidence value is available, - Compressing the respective quantized Quality markers so that the quality markers are compressed in an optimized manner.

2. Method according to claim 1, characterized in that the evaluation model is a genotype matching model, wherein in particular an uncertainty profile for a sequence of genome positions (121, 141) is determined.

3. Method according to claim 1 or 2, characterized in that the evaluation model is an activity-based tracking model, wherein in particular an activity profile for a sequence of genome positions (121, 141) is determined.

4. Method according to one of the preceding claims, characterized in that the determination of the confidence value of the respective quality marker (105) is carried out on the basis of a particularly weighted assignment of the respective quality marker (105) to a respective confidence interval.

5. Method according to one of the preceding claims, characterized in that the compression is carried out by means of quantizing the quality markers (105), in particular with subsequent entropy coding.

6. Method according to one of the preceding claims, characterized in that a comparison of several genome sequences (102, 104, 106, 108) is carried out to determine the confidence value, wherein the comparison of the genome sequences (102, 104, 106, 108) is carried out by means of an adjustment of respective Genome positions (121, 141) such that the respective genome sequences (102, 104, 106, 108) at respective aligned genome positions (121, 141) have respective identical genome positions (121, 141) of the genome sequence (102, 104, 106, 108).

7. Computer device with a compression unit which is configured to carry out a method according to one of the preceding claims.

Citation Information

Patent Citations

  • Efficient analog to digital converter

    CN105391450A

  • DNA sequencing quality score lossless compression method based on adaptive coding sequence

    CN111640467A

  • Bioinformatics systems, apparatuses, and methods executed on an integrated circuit processing platform

    US10691775B2

  • Electronic device and method for compressing sampled data

    US10938414B1

  • High-throughput DNA sequencing quality score lossless compression system and compression method

    WO2015180203A1