RU2818363C1 - Fast detection of gene fusions - Google Patents

Abstract

FIELD: biotechnology.

SUBSTANCE: described is a method of identifying one or more gene fusions in a biological sample. Method includes operations for obtaining first data, which represent a plurality of aligned readings, identifying a plurality of potential merges included in the obtained first data, filtering a plurality of potential fusions to determine a filtered set of potential merges, for each specific potential fusion from the filtered set of fusion candidates: generating input data by one or more computers for input into a machine learning model, which include extracted feature data representing a particular potential fusion, transmission of generated input data as input data to machine learning model, which has been trained to generate output reflecting the probability that the potential fusion is a confirmed gene fusion, and determining whether the particular potential fusion corresponds to the confirmed gene fusion based on the output. Also disclosed is a system for identifying one or more gene fusions in a biological sample, comprising one or more computers and one or more storage devices on which instructions are stored, when executed by one or more computers, one or more computers perform operations involving the described method. In addition, provided is a non-volatile machine-readable medium for identifying one or more gene fusions in a biological sample, on which software containing instructions is stored, made with possibility of execution by one or more computers, due to execution of which one or more computers perform operations, including described method.

EFFECT: fast detection of gene fusions.

13 cl, 4 dwg

Description

CROSS REFERENCE TO RELATED APPLICATION

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 62/944,304, filed December 5, 2019, which is incorporated herein by reference in its entirety.

PREREQUISITES FOR CREATION OF THE INVENTION

[0002] Gene fusions can be used as oncogenic drivers, which are important diagnostic and therapeutic targets in the treatment of diseases such as cancer.

SUMMARY OF THE INVENTION

[0003] In accordance with one innovative aspect of the present disclosure, a computer-implemented method for identifying one or more gene fusions in a biological sample is provided. In one aspect, a method may include the steps of one or more computers obtaining first data that represents a plurality of aligned reads from a read alignment block, one or more computers identifying a plurality of potential gene fusions included in the obtained first data, filtering the plurality of potential gene fusions by one or more computers to determine a filtered set of potential gene fusions, for each specific potential gene fusion from the filtered set of potential gene fusions: generating input data by one or more computers for input to a machine learning model, wherein generating such input data includes extracting characteristic data that distinguishes the particular potential fusion genes from data that includes: (i) one or more reference sequence segments to which a particular candidate gene fusion has been aligned using a read alignment block, and (ii) data generated from the output of the read alignment block, directed by one or more computers generated input data as input to a machine learning model, wherein the machine learning model has been trained to generate output data representing the probability that a potential gene fusion represents a confirmed gene fusion based on the results of applying the machine learning model to the input data being ( i) one or more reference sequence segments to which the particular candidate gene fusion was aligned using the read alignment block, and (ii) data generated based on the output of the read alignment block, receiving by one or more computers the output generated by the machine learning model after processing the generated input data by a machine learning model, and based on the output, determining whether a particular potential gene fusion is a confirmed candidate gene fusion by one or more computers.

[0004] Other versions include corresponding systems, apparatus, and computer programs for performing actions in a manner specified by instructions encoded on computer-readable storage devices.

[0005] These and other versions may optionally include one or more of the following features. For example, in some embodiments, generating the input data further includes retrieving feature data, which includes annotation data describing the annotations of the reference sequence segments to which the particular candidate gene fusion was aligned using the read alignment block. In such embodiments, the machine learning model is trained to generate output that reflects the probability that a candidate gene fusion is a confirmed candidate gene fusion based on the results of applying the machine learning model to the input data being: (i) one or more segments of a reference sequence, to which a particular potential gene fusion was aligned using a read alignment block, (ii) annotation data describing the annotations of reference sequence segments to which a particular potential gene fusion was aligned using a read alignment block, and (iii) data generated from the output of the block read alignment.

[0006] In some embodiments, identifying a plurality of potential gene fusions contained in the acquired first data by one or more computers may include identifying a plurality of split read alignments by one or more computers.

[0007] In some embodiments, identifying a plurality of potential gene fusions contained in the acquired first data by one or more computers may include identifying a plurality of alignments of discordant read pairs by one or more computers.

[0008] In some embodiments, a read alignment unit is implemented using a set of one or more processing modules that are configured to use hardware logic circuits physically arranged to perform operations using the hardware logic circuitry to: (i) receive data, representing the first read, (ii) matching the data representing the first read to one or more reference sequence regions to identify one or more matching reference sequence positions, (iii) generating one or more alignment estimates corresponding to each of the matching reference sequence positions for the first reading, (iv) selecting one or more candidate alignments for the first read based on the one or more alignment scores, and (v) outputting data representing the candidate alignment for the first read.

[0009] In some embodiments, a read alignment unit is implemented using a set of one or more processing modules by using one or more central processing units (CPUs) or one or more graphics processing units (GPUs) to execute software instructions that cause execution of one or more The CPU or one or more GPUs functions to: (i) receive data representing the first read, (ii) compare the data representing the first read to one or more reference sequence locations to identify one or more matching reference sequence positions for the first read, (iii ) generating one or more alignment estimates corresponding to each of the matching positions of the reference sequence for the first read, (iv) selecting one or more potential alignments for the first read based on the one or more alignment estimates, and (v) outputting data representing the potential alignment for the first reading.

[0010] In some embodiments, the method may further include the read alignment unit obtaining a plurality of reads that have not yet been aligned, the read alignment unit aligning a first subset of the plurality of reads, and the read alignment unit storing the first subset of the aligned reads in a storage device. In such embodiments, obtaining by one or more computers first data that represents a plurality of aligned reads from a read alignment unit may include obtaining by one or more computers a first subset of aligned reads from a storage device and performing one or more of the operations of claim 1, while how the read alignment unit aligns a second subset of the many reads that have not yet been aligned.

[0011] In some embodiments, the data generated based on the read alignment block output may include any one or more of variant allele frequency, number of unique read alignments, transcript read coverage, MAPQ score, or data indicating homology between parents genes.

[0012] In some embodiments, determining whether a particular candidate fusion matches a confirmed candidate gene fusion based on the output may include determining whether the output satisfies a predetermined threshold by one or more computers and determining that the particular candidate fusion matches the confirmed potential gene fusion, based on a determination that the output meets predefined thresholds.

[0013] In some embodiments, determining whether a particular candidate fusion matches a confirmed candidate gene fusion based on the output may include determining whether the output meets a predetermined threshold by one or more computers and determining whether the particular candidate fusion does not match the confirmed candidate gene fusion based on determining whether the output data does not meet predefined threshold values.

[0014] These and other innovative aspects of the present specification will be apparent from the detailed description, accompanying drawings, and claims.

BRIEF DESCRIPTION OF GRAPHIC MATERIALS

[0015] In FIG. Figure 1 shows a block diagram of an example system for rapid detection of confirmed gene fusions.

[0016] In FIG. 2 shows the flow of an example process for performing rapid detection of confirmed gene fusions.

[0017] In FIG. 3 shows a block diagram of another example of a system for rapidly detecting confirmed gene fusions.

[0018] In FIG. Figure 4 shows a block diagram of system components that can be used to implement a system for rapid detection of confirmed gene fusions.

DETAILED DESCRIPTION

[0019] The present description relates to systems, methods, devices, computer programs or any combination thereof for rapid detection of gene fusions. The presence of certain gene fusions may be important indicators of a particular disease, an indicator that suggests the use of a particular drug for a particular disease, or a combination of both. For example, certain gene fusions may be indicative of a particular form of cancer, such as acute and chronic myeloid leukemias, myelodysplastic syndromes (MDS), soft tissue sarcomas, or their treatments. The present disclosure allows for the rapid identification of accurate gene fusions by using a filtering module to reduce the number of potential gene fusions (also referred to herein as “potential fusions”), followed by processing to determine whether each potential fusion is a confirmed gene fusion. This filtering module allows for highly accurate selection of potential fusions for subsequent analysis while reducing the computational resources required to identify confirmed gene fusions, since only a filtered subset of potential gene fusions can be used for further processing, as described herein.

[0020] The reduced set of potential gene fusions also provides other technological advantages. For example, the methods and systems described herein provide reduced runtimes compared to standard methods for processing and evaluating all potential gene fusions. Reducing the runtime to perform operations also directly translates into lower costs in processing resources (such as CPU or GPU resources), storage usage, and power consumption. Although the filtration module provides reduced run time compared to traditional methods, the methods and systems described herein are also capable of using other approaches to reduce run time. For example, in some embodiments, further reductions in run time can be achieved by using a hardware-accelerated read alignment unit to map, align, and generate metadata used to process potential gene fusions.

[0021] In FIG. 1 shows a block diagram of an example system 100 for rapidly detecting confirmed gene fusions. The system 100 may include a nucleic acid sequencer 110, a storage device 120, a secondary analysis unit 130, a potential fusion identification module 140, a potential fusion filtering module 150, a feature set generation module 160, a machine learning model 170, a gene fusion detection module 180, an output module 190 application program interface (API) and output display 195. In the example shown in FIG. 1, each of these components is described as being implemented within the nucleic acid sequencer 110. However, the present description is not limited to such embodiments.

[0022] Instead, in some embodiments, one or more of the components shown in FIG. 1 can be implemented on a computer outside of the nucleic acid sequencer 110. For example, in some embodiments, secondary analysis modules may be implemented within a nucleic acid sequencer 110 and a potential fusion identification module 140, a potential fusion filtering module 150, a feature set generation module 160, a machine learning model 170, a gene fusion detection module 180, a module 190 output application programming interface (API), which can be implemented on one or more different computers. In such embodiments, one or more different computers and the nucleic acid sequencer may be communicatively coupled using one or more wired networks, one or more wireless networks, or a combination thereof.

[0023] For purposes of this specification, the term “module” includes one or more software components, one or more hardware components, or any combination thereof that can be used to implement functionality related to the corresponding module in this specification. Typically, the term "module" as used herein uses one or more processors to execute software instructions to implement the functionality of the module described herein. The processor may include a central processing unit (CPU), a graphics processing unit (GPU), or the like.

[0024] Likewise, as used herein, the term “block” includes one or more software components, one or more hardware components, or any combination thereof that can be used to implement functionality related to the corresponding block herein. Typically, a “block” described herein uses one or more hardware components, such as hardwired digital logic gates or hardwired digital logic blocks, organized as processing modules, to perform operations that implement the functionality of the block described herein. Such hardwired digital gates or hardwired digital logic circuits may include a field programmable gate array (FPGA), an application integrated circuit (ASIC), or the like.

[0025] Nucleic acid sequencer 110 (also referred to herein as sequencer 110) is configured to perform primary nucleotide sequence analysis. Performing a primary analysis may include introducing a biological sample 105, such as a blood sample, tissue sample, sputum, or nucleic acid sample, into the sequencer 110, and generating an output from the sequencer 110, such as one or more reads 112, each of which represents the nucleotide order of the nucleotide sequence. sequence of the obtained biological sample. In some embodiments, sequencing by the nucleic acid sequencer 110 can be performed over multiple read cycles, with a first "Read 1" cycle generating one or more first reads representing nucleotide order from the first end of the nucleotide sequence fragment, and a second "Read 1" cycle generating "Read 2" reads generate one or more second reads, respectively, representing the order of nucleotides from the other ends of one of the nucleotide sequence fragments. In some embodiments, the reads may be short reads of approximately 80 to 120 nucleotides in length. However, the present description is not limited to reads with any particular number of nucleotides. In contrast, the present disclosure can be used to read any number of nucleotides.

[0026] In some embodiments, biological sample 105 may include a DNA sample and nucleic acid sequencer 110 may include a DNA sequencer. In such embodiments, the order of sequenced nucleotides in the read produced by the nucleic acid sequencer may include one or more of guanine (G), cytosine (C), adenine (A), and thymine (T) in any combination. In some embodiments, a nucleic acid sequencer 110 may be used to obtain RNA reads from a biological sample 105. In such embodiments, this may occur using RNA-seq protocols. As an example, biological sample 105 may be preprocessed using reverse transcription to form complementary DNA (cDNA) using the enzyme reverse transcriptase. In other embodiments, the nucleic acid sequencer 110 may include an RNA sequencer, and the biological sample may include an RNA sample. RNA reads produced using cDNA or using an RNA sequencer can consist of C, G, A, and uracil (U). The example described herein in FIG. 1 is provided with reference to the generation and analysis of RNA reads. However, the present disclosure can be used to obtain and analyze any type of nucleotide sequence reads, including DNA or RNA reads.

[0027] Sequencer 110 may include a next generation sequencer (NGS) that is configured to generate sequence reads, such as reads 112-1, 112-2, 112-n, where "n" is any positive integer greater than 0. for a given sample in a manner that provides ultra-high throughput, scalability, and speed through the use of massively parallel sequencing technology. NGS allows the rapid sequencing of entire genomes, the ability to study sequenced target regions in depth, the use of RNA sequencing (RNA-Seq) to discover new RNA variants and splice sites or quantify mRNA for gene expression analysis, and the analysis of epigenetic factors such as DNA methylation in genomic scales and DNA-protein interactions, sequencing tumor samples to investigate rare somatic variants and tumor subclones, and studying microbial diversity, for example in humans or in the environment.

[0028] The sequencer 110 may sequence the biological sample 105 and generate a corresponding set of reads, represented as a set of A, C, T, and G. The sequencer may then perform reverse transcription to construct a cDNA sequence reflecting the corresponding RNA sequence. Such RNA sequence reads 112-1, 112-2, 112-n are output by sequencer 110 and stored in memory 120. In some embodiments, RNA sequence reads 112-1, 112-2, 112-n can be compressed into a smaller data record form volume before storing the reads 112-1, 112-2, 112-n in the storage device 120. The storage device 120 can be accessed by each of those shown in FIG. 1 components, including a secondary analysis block 130, a potential fusion identification module 140, a potential fusion filtering module 150, a feature set generation module 160, a machine learning model 170, a gene fusion detection module 180, and an output API module 190. Although the respective modules may be shown to provide output data from the first module to the second module, the practice of such functionality may involve the first module storing the output data in a storage device, such as memory device 120, and having the second module access the stored output data. in a storage device and processing the available output data as input data to the second module.

[0029] The secondary analysis unit 130 may access the reads 112-1, 112-2, 112-n stored in the storage device 120 and perform one or more secondary analysis operations on the reads 112-1, 112-2, 112-n . In some embodiments, reads 112-1, 112-2, 112-n may be stored in storage device 120 in compressed data records. In such embodiments, the secondary parsing unit may perform decompression operations on the compressed read records before performing secondary parsing operations on the read records. Secondary analysis operations may include mapping one or more reads to a reference genome, aligning one or more reads to a reference genome, or both. In some embodiments, the secondary analysis operations may also include variant recognition operations. In addition to performing secondary analysis operations, the secondary analysis unit 130 may also be configured to perform sorting operations. The sorting operations may include, for example, ordering the reads that have been aligned by the secondary analysis unit based on the positions in the reference genome to which the aligned reads have been mapped.

[0030] In some embodiments, such as the example shown in FIG. 1, the secondary analysis unit 130 may include a memory 132 and a programmable logic device 134. The programmable logic device 134 may include hardware logic circuits that can be dynamically configured to include one or more secondary analysis operational units, such as a read alignment unit 136, and can be used to perform one or more secondary analysis operations using hardware logic circuits. Dynamically configuring the programmable logic device 134 to enable a secondary analysis operation block, such as a read aligner 136, may involve, for example, transmitting one or more commands to the programmable logic device 134 that cause the programmable logic device 134's hardware logic elements to be translated by the programmable logic device 134 into a hardware digital logic configuration that is configured to implement the functionality of the read aligner 136 in the hardware logic.

[0031] The one or more operations resulting in a dynamic configuration of the programmable logic device 134 may include compiled hardware description language code, one or more instructions to the programmable logic device 134 to configure itself based on the compiled hardware description language code, or the like. P. Such operations, which result in a dynamic configuration of the programmable logic device 134, can be generated and transmitted to the programmable logic device 134 by a control program executed by the sequencer 110 or other computer hosting the control program. In some embodiments, the control program may be a software module whose instructions are stored in a storage device, such as memory device 120. The functionality of the control program to generate and transmit hardware description language instruction code or other instructions for configuring the programmable logic device 134 may be implemented by executing a control program software module using one or more processors, such as one or more CPUs or one or more GPUs.

[0032] The functionality of the read alignment block 136 may include retrieving one or more first reads, such as RNA reads 112-1, 112-2, 112-n, that were stored in the memory 120 of the sequencer 110, mapping the resulting first reads 112-1 , 112-2, 112-n with one or more reference sequence positions and then aligning the mapped first reads 112-1, 112-2, 112-n with the reference sequence. Thus, at the mapping stage, it is possible to identify a set of potential reference sequence positions for each specific read from the obtained first reads that correspond to a specific read. The alignment step can then generate a score for each potential reference sequence position and select the specific reference sequence position with the highest alignment score as the correct alignment for a particular read. The reference sequence may include an organized sequence of nucleotides corresponding to a known genome.

[0033] Configuring the hardware logic gates of programmable logic device 134 in response to one or more control program commands may include configuring logic gates, such as AND gates, OR gates, NOR gates, XOR gates, or any combination thereof, to perform digital logic functions of the block 136 read alignment. In an alternative or additional embodiment, the hardware logic element configuration may include dynamically configurable logic blocks comprising configurable hardware logic blocks to perform complex computational operations, including addition, multiplication, comparison, or the like. The exact configuration of hardware logic elements, logic blocks, or combinations thereof is determined by commands received from the control program. The received instructions may include (or be derived from) compiled hardware description language (HDL) program code that has been written by the subject and specifies a secondary analysis operating unit layout design to be programmed into the programmable logic device 134. The HDL program code may include software code written in a language such as high-speed integrated circuit hardware description language (VHDL), Verilog, or the like. The subject may be one or more human users who generated the HDL code, one or more artificial intelligence agents that generated the HDL code, or a combination of both.

[0034] Programmable logic device 134 may include any type of programmable logic device. For example, programmable logic device 134 may include one or more field programmable gate arrays (FPGAs), one or more complex programmable gate arrays (CPLDs), or one or more programmable gate arrays (PLAs), or a combination thereof, that are dynamically configurable and reconfiguration by the control program as necessary to perform a specific sequence of operating operations. For example, in some implementations it may be desirable to use programmable logic device 134 as read aligner 136, as described above. However, in other embodiments, it may be desirable to use programmable logic device 134 to perform variant recognition functions or variant recognition supporting functions, such as a Hidden Markov Model (HMM) block. In other implementations, the programmable logic device 134 may also be dynamically configured to support common computing tasks such as compression and decompression, since the hardware logic of the programmable logic device 134 is capable of performing these tasks and other tasks noted above much faster than performing the same tasks using software instructions executed by one or more processors 150. In some embodiments, programmable logic device 134 may be dynamically reconfigured during a run to perform various operations.

[0035] As an example, in some embodiments, programmable logic device 134 may be implemented using an FPGA that is dynamically configured as a decompressor to access data representing a compressed version of the first reads 112-1, 112-2, 112-n. stored in memory 120 or 132. Secondary analysis unit 130 may use a decompressor to decompress compressed data representing first reads 112-1, 112-2, 122-n (e.g., if reads obtained from a nucleic acid sequencer have been compressed) . The decompression unit may store the decompressed reads in memory 120 or 132. In such embodiments, the FPGA may then be dynamically reconfigured as a read alignment unit 136 and used to perform mapping and alignment of the decompressed first reads 112-1, 112-2, 112-n already stored in memory 132 or 120. Read aligner 136 may then store data representing the mapped and aligned reads in memory 132 or 120. Although such a sequence of operations is described as including decompression, mapping, and alignment operations, the present description is not limited to performing these operations or only these operations. In contrast, programmable logic device 134 may be dynamically configured to implement the functionality of any operating block, as needed, in any order to implement the functionality described herein.

[0036] In the example of FIG. 1, a secondary analysis block 130 is described that uses hardware logic in the form of a programmable logic device 134 to implement a read alignment block 136. However, the present description is not limited to the use of programmable logic devices to implement the read alignment unit 136. In contrast, other types of integrated circuits may be used to implement the read alignment block 136 in the hardware digital logic circuit of the secondary analysis block 130. For example, in some embodiments, the secondary analysis block 143 may be configured to use one or more application specific integrated circuits (ASICs) to implement the functionality of one or more secondary analysis operational blocks. Although not reprogrammable, one or more application specific integrated circuits (ASICs) may be constructed using dedicated hardware logic of one or more secondary analysis operating units, such as a read alignment unit 136, a variant recognition unit, a variant recognition support computational unit, or the like. ., to speed up and parallelize the execution of secondary analysis operations. In some implementations, the use of one or more ASICs as hardware logic circuits of the secondary analysis block 130 that implement the functionality of one or more secondary analysis blocks may provide even higher speed than the use of a programmable logic device such as an FPGA. Accordingly, one skilled in the art will appreciate that in any of the embodiments described herein, an ASIC may be used in place of a programmable logic device such as an FPGA. For implementations that will use ASICs, it is necessary to use a dedicated ASIC or dedicated logical groups of a single ASIC for each block of secondary analysis operations that the ASIC must perform. By way of example, one or more read alignment ASICs, one or more decompression ASICs, one or more compression ASICs, or combinations thereof are used. Additionally, the same functionality can also be provided by using dedicated logic groups in the same ASIC.

[0037] In addition, in the examples of the present description presented with reference to systems 100 and 300 in FIG. 1 and 3, respectively, are described with reference to the use of a hardware implementation of a read alignment unit 136 in a programmable logic device. Additionally, as noted above, one or more ASICs may be used to implement the read alignment module or other secondary analysis blocks. However, the present description is not limited to the use of hardware units for implementing such secondary analysis operations. Conversely, in some embodiments, any of the operations described herein performed by a programmable logic device, such as read alignment, compression, or decompression, may also be implemented using one or more software modules.

[0038] As shown in the example of FIG. 1, an implementation of system 100 may begin with a sequencer 110 to sequence a biological sample 105. Sequencing a biological sample may involve the sequencer 110 generating sequence reads, which are data reflecting the ordered sequences of nucleotides present in the biological sample 105. If the system 100 is configured to process DNA reads, reads generated by sequencer 110 may be stored in memory 120.

[0039] Alternatively, in some embodiments, if the system 100 is configured to process RNA reads, the sequencer 110 may be configured to preprocess the biological sample 110 using reverse transcription to generate complementary DNA (cDNA) using a reverse enzyme transcriptase. In such embodiments, such as the example embodiment shown in FIG. 1, the reads generated by the sequencer 110 include RNA reads 112-1, 112-2, 112-n. In other embodiments, the nucleic acid sequencer 110 may include an RNA sequencer, and the biological sample may include an RNA sample. Whether RNA reads are generated by a DNA sequencer using cDNA or by an RNA sequencer, each of the RNA reads includes a nucleotide sequence consisting of C, G, A and U. Reads 112-1, 112-2, 112- n may be stored in storage device 120 in a compressed or decompressed format.

[0040] The implementation of system 100 may continue with secondary analysis unit 130 receiving reads 112-1, 112-2, 112-n stored in storage device 120. In some embodiments, secondary analysis unit 130 may be given access to reads 112 -1, 112-2, 122-n in the memory 120, and it can store the available reads 112-1, 112-2, 112-n in the memory 132 of the secondary analysis unit 130. In other embodiments, after the manager registers the completion of sequencing of reads 112-1, 112-2, 112-n and the availability of secondary analysis unit 130 to perform secondary analysis operations, the manager may load reads 112-1, 112-2, 112-n into memory 132 of the secondary analysis unit 130.

[0041] In the case of compression of reads 112-1, 112-2, 112-n, secondary analysis unit 130 may dynamically configure programmable logic device 134 as a decompressor to access reads 112-1, 112-2, 112-n in the storage device 132 or 120, decompress reads 112-1, 112-2, 112-n, and then store the decompressed reads 112-1, 112-2, 112-n in storage device 1320 or 120. In some embodiments, the secondary analysis unit may dynamically change the configuration of a programmable logical device and perform unpacking in accordance with the commands of the control program.

[0042] If reads 112-1, 112-2, 122-n have not been compressed, secondary analysis unit 130 may access reads from memory 132 or 120 and perform read alignment operations. In some embodiments, the secondary analysis unit 130 may receive a command from the control program instructing the secondary analysis unit 130 to configure or reconfigure the programmable logic device 134 to enable the read alignment unit 136, and then use the read alignment unit 136 to perform alignment of the reads 112- 1, 112-2, 112-n. Alternatively, in other embodiments, the programmable logic device may already be configured to include a read alignment block 136 and use the read alignment block 136 to perform read alignment 112-1, 112-2, 112-n. In other embodiments, the secondary analysis unit 130 may include an ASIC that is configured to perform read alignment and then use the ASIC to perform read alignment 112-1, 112-2, 112-n.

[0043] The secondary analysis unit 130 may be configured to perform read alignment operations in parallel with the gene fusion analysis. For example, secondary analysis unit 140 may receive a first batch of reads generated by sequencer 110 that have not been aligned, use read alignment unit 136 to align the first batch of reads, use a sort module that may be implemented in the hardware configuration of programmable logic device 136, or may implemented in software by executing software instructions to sort the aligned reads and then output the first batch of aligned and sorted reads for storage in memory 132, 130. In some embodiments, memory 132 may function as a local cache for secondary analysis unit 132, into which loading data to be processed in the read alignment unit, and then unloading the data that is output by the read alignment unit 136. Therefore, after the first batch of aligned reads is output from the read alignment unit 136 to memory 132, the first batch of aligned reads may be sorted and then loaded into memory 120. The potential merge identification module 140 may then access the first batch of aligned and sorted reads from the storage device. 120 and begin processing the first batch of aligned and sorted reads while the secondary analysis unit 130 performs alignment operations on the second batch of reads that were prepared by sequencer 110 and were not previously aligned. Such a process can be performed iteratively until the system 100 has processed each batch of reads. Although this example is described using aligned and sorted packets, the present description does not require that the aligned read packets also be sorted. In contrast, the use of aligned and sorted reads can be implemented in system 100 or system 300 to improve performance, such as reducing run time, as described below.

[0044] Potential fusion identification module 140 may receive a batch of aligned and sorted reads that have been aligned by read alignment unit 136 and determine whether the batch of aligned and sorted reads contains one or more potential fusion genes. In some embodiments, if the resulting package includes aligned and sorted reads, the potential fusion identification module 140 may evaluate the sorted reads of the package, wherein the genomic interval corresponding to the package overlaps the breakpoint of at least one potential fusion. This may reduce the number of potential mergers that require further analysis. In other implementations, if a received packet includes aligned reads that have not been sorted, the potential merge identification module 140 may evaluate each of the aligned reads in the packet to determine whether the aligned read is a potential merge. In some embodiments, the operation of determining whether a package of reads includes one or more potential fusions by the potential fusion identification module 140 involves the identification of a potential fusion by the module 140, where the package of reads includes one or more split read alignments, one or more discordant pairs reads, one or more limited-clipping alignments, or a combination thereof.

[0045] In some embodiments, potential fusion identification module 140 may be configured to identify split read alignments as potential fusions. Potential fusion identification module 140 can identify split read alignments by analyzing the reference sequence genes to which each particular read in the stack of aligned reads was aligned. If the potential fusion identification module 140 identifies a read as a single gene, the potential fusion identification module 140 may determine that the read is not split. In an alternative embodiment, if the potential fusion identification module 140 determines that a read matches two different genes, then the read may be considered split. In such embodiments, split reads may be classified as potential fusions. A read can be considered to correspond to two different reads if, for example, a first subset of read nucleotides is aligned to a first parent gene of a reference genome and a second subset of read nucleotides is aligned to a second parent gene of a reference genome. In some embodiments, the first subset of nucleotides may be a read prefix and the second subset of nucleotides may be a read suffix. If the potential fusion identification module 140 is configured to identify split reads, split read identification data, if any, may be stored in memory 120.

[0046] In some embodiments, potential fusion identification module 140 may be configured to identify alignments of discordant read pairs as potential fusions. Potential fusion identification module 140 may identify discordant read pairs by analyzing the genes of the reference sequence to which each particular read pair in the batch of aligned reads was aligned. If a read pair is aligned to a reference sequence and the orientation and range of the alignment matches the expected orientation and range, the read pair is not a discordant read. In an alternative embodiment, if a read pair is aligned to a reference sequence and the orientation and range of the alignment matches the expected orientation and range, the read pair will be classified as discordant reads. In such embodiments, if one read in a pair corresponds to one parent gene and the other corresponds to the other parent gene, the discordant read may be classified as a potential fusion. If the potential merger identification module 140 is configured to identify discordant reads, the discordant read identification data, if any, may be stored in the storage device 120.

[0047] In some embodiments, potential merge identification module 140 may be configured to identify clipping-constrained alignments as potential merges. Potential fusion identification module 140 can identify constrained clipping alignments by analyzing the reference sequence genes to which each particular read in the batch of aligned reads was aligned. In some embodiments, the potential fusion identification module 140 may determine whether a read is fully aligned to a single position in the reference genome. If the potential fusion identification module 140 identifies a complete alignment of a particular read to a single position in the reference genome, the potential fusion identification module 140 may determine that the read is not a split read. In an alternative embodiment, if the potential fusion identification module 140 determines that only a portion of a particular read aligns with the reference genome, the potential fusion identification module 140 may determine that the read is split. If the aligned read region corresponds to one parent gene, and the unaligned region contains sequence similar to the other parent gene, the clipping-constrained read will be classified as a potential fusion. If the potential fusion identification module 140 is configured to identify clipping-constrained reads, data identifying the clipping-constrained reads as candidate fusions, if any, may be stored in memory 120.

[0048] The potential merger filtering module 150 may obtain data describing the set of potential mergers identified by the potential merger identification module 140 . In some embodiments, a potential merger filtering module may access memory 120 and obtain data describing potential mergers from memory 120. In other embodiments, a potential merger filtering module may obtain data describing potential mergers from the output of a previous module. such as potential merger identification module 140. The potential fusion filtering module 150 may use one or more filters to filter data describing the set of potential fusions to identify a filtered set of potential gene fusions that is less than the complete set of potential fusions. In some embodiments, such filters are applied within a single stage. For example, each of one or more filters may be applied, and each candidate merger in a set of candidate mergers may be evaluated after each of the one or more filters. However, in other embodiments, multi-stage filtering approaches may be used. In such embodiments, a first set of one or more filters is applied to the initial set of candidate mergers identified by the potential merger identification module 140 . A second set of one or more filters is then applied to the first set of filtered merge candidates that remain after the first filtering step. If necessary, additional filtering stages can also be applied to obtain the optimal filtered set of potential candidates.

[0049] In some embodiments, fusion candidate filtering module 150 may filter the set of fusion candidates to account for redundant fusion candidates that result from high depth of coverage in short read sequencing. For example, the accumulation of data from 30-fold sequencing may result in the potential fusion identification module 140 identifying up to 30 potential fusions that are duplicates of each other. The potential merger filtering module 150 may remove such duplicate potential mergers by applying a filter to the parameters of the potential mergers to identify duplications. For example, potential fusion filtering module 150 may determine whether a plurality of potential fusions are aligned to the same parent gene, whether they are aligned to a region of a reference genome that includes the same or a similar point break, or a combination of both. If the potential fusion filtering module 150 identifies a plurality of potential fusions that are aligned to the same parent gene, aligned to a region of a reference genome that includes the same or a similar point break, or a combination thereof, the potential fusion filtering module 150 may determine whether they are whether the potential mergers are duplicates, and select only one potential merger as a representative potential merger. In such cases, the remaining potential fusions that were aligned to the same parent gene, aligned to a region of the reference genome that included the same or a similar point break, or a combination of both, can be excluded without further downstream analysis. The representative candidate merger can then be added to the set of filtered candidate mergers in a storage device, such as storage device 120.

[0050] In an alternative or additional embodiment, merger candidate filtering module 150 may filter a set of merger candidates based on one or more rule-defined conditions. For example, the potential merger filtering module 150 may analyze each potential merger and determine whether the potential merger matches one or more parameters that satisfy one or more rule-defined conditions used by the filtering modules 150. In some embodiments, one or more rule-specified conditions may include the alignment position of each candidate merge region, the overlap distance of the alignment relative to the point gap covered by the potential merge, the orientation of the potential merge alignment, the quality of the read alignment of the potential merge, the additional alignment of the position of the potential merge, or any combination thereof. .

[0051] As an example, potential merge filtering module 150 may use one or more settable condition rules to filter potential merges based on alignment position. For example, in some embodiments, the potential fusion filtering module 150 may be configured to use a rules-based condition that filters out potential fusions with reads aligned to a reference sequence such that the coverage of the alignment crosses the fusion break point by more than a predetermined number of nucleotides. In some embodiments, the predetermined number of nucleotides for such a rule condition may be 8 nucleotides. In an alternative or additional embodiment, the potential fusion filtering module 150 may be configured to filter potential fusions with reads aligned to a reference sequence such that the coverage of the alignment on the reference sequence does not reach a predetermined limit of nucleotides from the fusion break point. In some embodiments, the predetermined threshold number of nucleotides for such a rule-driven condition may be 50 nucleotides. In an alternative or additional embodiment, the potential fusion filtering module 150 may be configured to use a rule-based condition that filters out potential fusions with reads aligned to a reference sequence such that the aligned reads at two point breaks share at least a predetermined number of nucleotides. In some embodiments, the predetermined number of total nucleotides may include at least 8 nucleotides.

[0052] As another example, potential merge filtering module 150 may use one or more rule-based conditions to filter potential merges based on orientation. In some embodiments, for example, the potential fusion filtering module 150 may be configured to apply a rules condition that filters out potential fusions with an alignment orientation indicating a reversal of the nucleotide sequence of at least one of the parent genes in the fusion transcript.

[0053] As another example, potential merger filtering module 150 may use one or more rule-based conditions to filter potential mergers based on mapping quality. For example, in some embodiments, the potential merger filtering module 150 may be configured to use a rule-based condition that filters out potential aligned read mergers with a mapping quality score that does not meet a predefined threshold.

[0054] As another example, potential merger filtering module 150 may use one or more rule-based conditions to filter potential mergers based on additional mapping provisions. In some embodiments, for example, the potential fusion filtering module 150 may be configured to apply a rule-based condition that filters out potential fusions based on determining that the candidate fusion read region matches a plurality of reference sequence positions. In some embodiments, the potential fusion filtering module 150 may be configured to exclude positions annotated as homologous genes.

[0055] Potential mergers that satisfy each of the one or more rule-defined conditions may be added to the set of filtered candidate mergers in a storage device, such as storage device 120. Potential mergers that do not satisfy each of the one or more rule-defined conditions may be excluded without further follow-up analysis. In some embodiments, a rule-based condition based on the results of filtering potential mergers may be applied as a second-stage filter after the first-stage deduplication filter has been applied. In other embodiments, filtering potential mergers based on a condition specified by the rules may be applied as a first stage of filtering, and then a deduplication filter may be applied as a second stage filter. In other embodiments, filtering based on a rule condition may be applied as a single filtering step without prior deduplication filtering. Filtering potential mergers based on one or more of these rule-based conditions can significantly reduce the number of potential mergers that will need to undergo additional post-processing.

[0056] Subsequent processing may be performed for each candidate merger in the filtered set of candidate mergers output by the candidate merger filtering module 150 . Subsequent processing includes performing the operations of a feature set generation module 160, a machine learning model 170, a gene fusion detection module 180, and an output API module 190. Such post-processing can be used to determine whether a candidate fusion matches a confirmed candidate gene fusion.

[0057] The feature set generation module 160 may rely on data from a plurality of data sources to identify a set of data attributes on which to base feature extraction. Such data sources include attribute data stored in memory 120 about the candidate fusion, which includes (i) the candidate fusion read(s), (ii) the region(s) of reference sequence positions to which the candidate fusion reads were aligned, and (iii) annotations of the reference genome segments to which a particular candidate alignment was aligned. In some embodiments, the annotations may include annotations of gene exons, annotations indicating the presence of homologous genes, annotations indicating a list of enriched genes, or a combination thereof.

[0058] The data sources that feature set generation module 160 may use may also include data generated by read alignment unit 136 during the alignment process. In some embodiments, feature set generation module 160 may obtain feature data based on data generated by read aligner 136 when aligning a potential fusion. For example, the feature set generation module 160, based on the data generated by the read alignment block 136, may obtain information such as the frequency of occurrence of variant alleles, the number of unique read alignments, transcript read coverage, MAPQ score, data indicating homology between parent genes, or their combination.

[0059] A feature set generation module 160 may be used to generate feature data that represents one or more of the aforementioned potential fusion attributes obtained from a plurality of data sources, and encode the feature data into one or more data structures 162 for input to a machine learning model 170 . For example, in some implementations, the entire set of features derived from the attributes of a potential merge may be encoded into a unit vector 162 included in the machine learning module 170 . For example, in a split-read or constrained-clipping alignment scenario, each of the features derived from the attributes of these kinds of potential fusions could be encoded in unit vectors 162.

[0060] In other embodiments, the feature data may be derived from attributes of potential mergers and may be encoded for input as a plurality of vectors. In such a scenario, input vector 162 may include a pair of input vectors 162a, 162b. For example, in a split-read scenario of a potential fusion, each of the features extracted from the attributes related to the split-read prefix, including features representing the nucleotides of the split-read prefix, features representing the segment of the reference sequence to which the prefix is aligned, and any other features extracted from the above prefix-related attributes, or any combination thereof, may be encoded in input vector 162a. Likewise, in such an embodiment, each of the features extracted from the attributes related to the split-read suffix, including features representing the nucleotides of the split-read prefix, features representing the segment of the reference sequence to which the suffix is aligned, and any other features extracted from the above attributes , related to the suffix, or any combination thereof may be encoded in the input vector 162b. As another example, if a discordant read pair is identified as a potential fusion, extracted features representing the first read of the discordant read pair, extracted features representing the region of the reference sequence to which they were aligned, features extracted from attributes related to the first read of the discordant pair reads, or any combination thereof, may be encoded in input vector 162a. Similarly, in such an example, extracted features representing the second read of the discordant read pair, extracted features representing the region of the reference sequence to which they were aligned, features extracted from attributes relating to the second read of the discordant read pair, or any combination thereof, may be encoded in the input vector 162b.

[0061] Each of one or more vectors 162 may numerically represent generated feature data, such feature data including any of the features obtained from potential mergers or any of the features obtained from read alignment block 136 data related to a potential merger and stored in memory 120. For example, each vector 162 or 162a, 162b may include a plurality of fields, each corresponding to a particular feature of a particular read of a particular candidate fusion. Depending on the specific potential merger, this may result in one or more input vectors, as described above. The feature set generation module 160 may determine a numerical value for each of the fields that describes the degree to which a particular feature appears in the attributes of a particular candidate fusion read. The established numeric values for each of the fields can be used to encode the prepared feature data reflecting the attributes of the potential fusion reads into one or more corresponding vectors 162. The constructed one or more vectors 162a, 162b that numerically reflect the corresponding candidate fusion reads are input as input information into the machine learning model 170. In some implementations, even if multiple concept vectors are generated for a given candidate merger, the multiple concept vectors can be collapsed into a single vector 162 that can be input to a machine learning model 170 . In such implementations, if it was practical to obtain multiple vectors in (i) certain split-read implementations with prefix features assigned to the first vector and suffix features assigned to the second vector, or (ii) in discordant pair implementations, then the first portion of the unit vector may correspond to the conceptual first vector, and the second portion of the unit vector may correspond to the conceptual second vector.

[0062] The machine learning model 170 may include a deep neural network trained to calculate the probability that a potential fusion matches a confirmed gene fusion based on processing the input information of one or more input vectors 162 that reflect features of a particular potential fusion. A confirmed gene fusion is a chimeric transcript that contains a sequence from multiple genes due to a rearrangement in the genome, linking the prefix of one parent gene to the suffix of another parent gene. As used herein, a confirmed gene fusion will be considered predicted by the model 170 if, for example, the output 178 produced by the machine learning model satisfies a predefined threshold. The machine learning model 170 may include an input layer 172 for receiving input data, one or more hidden layers 174a, 174b, 174c for processing input data received through the input layer 172, and an output layer 176 for providing output data 178. Each hidden layer 174a, 174b, 174c includes one or more weighting factors or other parameters. The weights or other parameters of each respective hidden layer 174a, 174b, 174c may be adjusted during training such that the trained deep neural network produces the desired target output 178 indicating the probability that one or more input vectors 162 represent a confirmed gene fusion based on a machine learning model 170 processing one or more input vectors 162.

[0063] The machine learning model 170 can be trained in several different ways. In one implementation, a machine learning model 170 may be trained to distinguish between (i) one or more input vectors containing features derived from attributes of confirmed potential mergers and (ii) one or more input vectors containing features derived from attributes of unconfirmed potential mergers. In some embodiments, such training may be performed using labeled pairs of training vectors. Each training vector may represent a training candidate fusion and may consist of the same types of feature data as one or more of the input vectors 162 mentioned above. In such embodiments, one or more input vectors 162 containing features derived from attributes of potential fusions may be labeled as confirmed gene fusions or unconfirmed gene fusions. In some embodiments, the confirmed gene fusion mark and the unconfirmed gene fusion mark may be represented as a numerical value. For example, in some embodiments, the label of a confirmed gene fusion may be "1" and the label of an unconfirmed gene fusion may be "0". In other embodiments, for example, the label of a confirmed gene fusion may be a number from "0" to "1" that meets a predefined threshold, and the label of an unconfirmed gene fusion may be a number from "0" to "1" that does not satisfy a predefined threshold. threshold value. In such embodiments, the value at which such number does or does not satisfy a predetermined threshold indicates the level of confidence that the training pair of input vectors represents a confirmed gene fusion or an unconfirmed gene fusion. In some embodiments, meeting a predetermined threshold may include exceeding a predetermined threshold. However, embodiments may also be configured such that meeting a threshold value means not exceeding a predetermined threshold value. Such implementations may include, for example, implementations that eliminate the use of a comparator and parameters.

[0064] In the training process, each labeled set of one or more training vectors is input to a machine learning model 170 for processing by the machine learning model 170, and then the training output prepared by the machine learning model 170 is used to determine a predicted label. for each labeled set of one or more training vectors. The predicted label obtained by the machine learning model 170 based on the machine learning model's processing of one or more labeled training vectors corresponding to a pair of reads of a training candidate fusion may be compared with a training label for one or more training vectors corresponding to one or more reads (or read sites) teaching potential merger. The parameters of the machine learning model 170 can then be adjusted based on the differences between the predicted labels and the training labels. This process may iteratively continue for each of the plurality of labeled training vectors corresponding to the corresponding training candidate merger, as long as the predicted potential merger labels obtained by the machine learning model 170 based on processing the set of one or more training vectors corresponding to the training candidate merger are within a predetermined error levels will not match the training labels of the set of one or more training vectors associated with the corresponding training candidate fusion.

[0065] In some embodiments, tagged training candidate mergers may be obtained from a library of training candidate mergers that have been analyzed and labeled by one or more human users. However, in other embodiments, the tagged training merge candidates may include a training merge candidate that was generated and tagged by the simulator. In such embodiments, the simulator can be used to generate distributions of various categories of training candidate mergers that can be used to train a machine learning model 170 . Typically, if a run of a machine learning model 170 involves inputting a single input vector 162 in which each extracted potential fusion feature is encoded by a single input vector 162, the machine learning model 170 must be trained using a single input vector with the same features as the input vector 162 ,using the learning process described above. Likewise, if a run of machine learning module 170 involves inputting two training vectors 162a, 162b as described above, then machine learning model 170 must be trained using two input vectors, each of which has the same corresponding features of the above input vectors 162a, 162b. Thus, the type of input vectors to be processed during a run is the same as the type of vectors to be used to train the model 170 using the training process described above.

[0066] While processing input data 162 that corresponds to features derived from attributes of a potential merge, the output of each hidden layer 174a, 174b, 174c may include an activation vector. The output activation vector output by each respective hidden layer can be propagated through subsequent layers of the deep neural network and used by the output layer to generate output 178. In the example shown in FIG. 1, machine learning model 170 is trained to generate output 178, which is a composite score obtained by machine learning model 170 based on the machine learning model's processing of individual input vectors 162a, 162b, each corresponding to one of the candidate fusion reads. Such a combined score 178 is ultimately generated by the output layer 176 of the trained machine learning model based on the calculations performed by the output layer 176 of the trained machine learning model 170 with the activation vector obtained from the final hidden layer 174c.

[0067] Gene fusion detection module 180 may evaluate output 178 produced by trained machine learning model 170 to determine whether it indicates that a candidate fusion corresponding to one or more input vectors 162 is a confirmed candidate fusion. In some implementations, the output 178 may be provided to the gene fusion detection module 180 by the trained machine learning model 170 . In other implementations, the system 100 may store the output 178 of the trained machine learning model 170 in a storage device, such as storage device 120, for later access by the gene fusion detection module 180. Gene fusion detection module 180 may receive output 178 prepared by machine learning model 170 and evaluate output 178 to determine, based on output 178, whether a potential fusion corresponding to a pair 162 of input vectors 162a, 162b is a confirmed gene fusion. In some embodiments, gene fusion determination module 180 may determine whether a potential fusion corresponding to one or more input vectors 162 is a confirmed gene fusion by comparing the output 178 produced by the machine learning model to a predefined threshold. If gene fusion determination module 180 determines that output 178 satisfies a predetermined threshold, then gene fusion determination module 180 may consider a potential fusion corresponding to one or more input vectors 162 to be a confirmed gene fusion. In an alternative embodiment, if gene fusion determination module 180 determines that output 178 does not meet a predetermined threshold, gene fusion determination module 180 may consider that a potential fusion corresponding to one or more input vectors 162 is not a confirmed gene fusion.

[0068] In some embodiments, gene fusion determination module 180 may generate output 182 that indicates the results of a decision made by gene fusion determination module 180 based on gene fusion determination module 180's assessment of output 178 produced by machine learning model 170. Such output data 182 may include potential gene fusion identification data that corresponds to one or more input vectors 162 and data containing the decision made by the gene fusion determination module 180 . Data containing the decision of gene fusion determination module 180 may include data indicating whether a potential gene fusion that matches one or more input vectors 162 is a confirmed gene fusion or an unconfirmed gene fusion. In some embodiments, output 182 may contain only a list of confirmed gene fusions identified based on output 178, a list of unconfirmed gene fusions identified based on output 178, data that no confirmed gene fusions were found, or any combination thereof. In some embodiments, such output data 182 may be stored in storage device 182 for later use by another computing module, for subsequent output to a user device, or the like.

[0069] In an alternative or additional embodiment, gene fusion detection module 180 may generate output data 184 that can be passed as input to output application program interface (API) module 190. Output 184 may include instructions for the output API to display an output indicating whether a potential gene fusion that matches one or more input vectors 162 is a confirmed gene fusion or an unconfirmed gene fusion. In some implementations, commands may instruct output API module 190 to access output data 182 stored in storage device 120 and generate transformed data that, when converted by a computing device associated with output display 195, instructs output display 195 to display (i) the data identifying potential fusions that correspond to one or more input vectors 162, and (ii) data indicating whether the identified potential fusion is a confirmed gene fusion or an unconfirmed gene fusion. In this case, the output display 195 may be instructed to display any output data 182 stored in the storage device 184. In some embodiments, such output information may be output in the form of a report.

[0070] In some embodiments, gene fusion detection module 180 stores output 182 for each candidate gene fusion in memory 120 from post-processing performed on each candidate fusion from the filtered set of candidate gene fusions. In such embodiments, gene fusion detection module 180 may only instruct output API module 190 to output gene fusion analysis results stored in memory 120 for each candidate fusion from the filtered set of candidate gene fusions after post-processing of each candidate fusion is completed. In such a scenario, the output 192 provided for display on the output display 195 will include a list of confirmed gene fusions, a list of unconfirmed gene fusions, or both lists. In other embodiments, the gene fusion detection module 180 may command the output API module 190 to output result data indicating a list of identified gene fusions, if any, after further processing is completed for that particular candidate fusion.

[0071] APIR output module 190 may output output data 192 in other forms. For example, in some embodiments, output data 192 may be data that outputs a report by another device, such as a printer, including (i) data identifying a potential merger that corresponds to one or more vectors 162, and (ii) data indicating whether the identified potential fusion is a confirmed gene. In other embodiments, output 192 may cause the speaker to output audio data including (i) data identifying a potential fusion that matches one or more vectors 162 and (ii) data indicating whether the identified potential fusion is a confirmed gene. APIR output module 190 may also output output in other forms.

[0072] In some embodiments, the output display 195 may be a display panel of the sequencer 110. In other embodiments, the output display 195 may be a display panel of a user device that is coupled to the sequencer 110 via one or more networks. Indeed, sequencer 110 can be used to transmit output 192 to any device with any display.

[0073] In FIG. 2 shows an example flow of a process 200 for conducting rapid detection of confirmed gene fusions. A system, such as system 100, may begin executing process 200 using one or more computers to receive first data representing a plurality of aligned reads from the read alignment unit (210). The system can identify many potential gene fusions included in the first data obtained (220). The system can filter multiple potential gene fusions to generate a filtered set of potential gene fusions (230).

[0074] The system may retrieve a specific candidate gene fusion from a filtered set of candidate gene fusions (240). The system may generate input data for input to a machine learning model, wherein generating the input data includes extracting representative data representing a particular candidate gene fusion from data comprising (i) one or more reference sequence segments to which the particular candidate gene fusion was aligned using a read alignment block, and (ii) data generated based on the output of the read alignment block (250).

[0075] The system may provide the prepared input data as input to a machine learning model, wherein the machine learning model is trained to generate an output reflecting the probability that a potential gene fusion is a confirmed gene fusion based on the results of applying the machine learning model to the processing input data representing (i) the reference genome segments to which a particular candidate gene fusion was aligned using the read alignment block, and (ii) data generated from the output of the read alignment block (260). The system may receive output generated by the machine learning model from the output of the machine learning model processing the input data (270). The system can determine whether a potential fusion is a confirmed gene fusion based on the output data (280).

[0076] After completion of step 280, the system may determine whether to evaluate another candidate merger from the filtered set of candidate mergers (290). If the system determines that there is another candidate merger from the evaluated filtered set of candidate mergers, the system may continue executing process 200 at step 240. In an alternative embodiment, if the system determines that there is no other candidate merger from the evaluated filtered set of potential mergers, the system may terminate the process at step 295. Another candidate merger may be present in the filtered set of candidate mergers if the set of candidate mergers has not been exhausted.

[0077] In FIG. 3 shows a block diagram of another example of a system 300 for rapidly detecting confirmed gene fusions. System 300 performs the same functions as system 100 in that system 300 uses sequencer 110 to generate RNA (or DNA) sequence reads 112, uses secondary analysis unit 130 to align RNA sequence reads 112 to a reference sequence, uses potential fusion identification module 140 to identify a potential fusion, uses the potential fusion filtering module 150 to generate a filtered set of potential fusions for further analysis, and then performs subsequent analysis on the filtered set of potential fusions to identify confirmed gene fusions using the feature set generation module 160, machine learning model 170, module 190 gene fusion definitions and output API module 190. Each of these functional blocks, modules or models performs the same functions that were described for them when considering the system 100 in FIG. 1.

[0078] The difference between system 300 and system 100 is that identification of potential fusions, filtering of potential fusions, and further analysis of the filtered set of candidate fusions are performed on another computer 320 rather than on sequencer 110. Accordingly, the differences between system 300 and system 100 consist of a method for packaging the aligned reads and transferring them to the gene fusion analysis computer 320 using the network 310, unpacking by the computer 320, and a method for packaging and transferring the results of the gene fusion analysis to another device with an associated display for output.

[0079] In more detail, sequencer 110 may sequence biological sample 105 and generate RNA reads 112-1, 112-2, 112-n, where “n” is any positive integer greater than 0, as described with reference to system 100. Although RNA reads are used as an example, the system can also perform the same procedures with DNA reads. Sequencer 110 may store reads 112-1, 112-2, 112-n in memory 120. In some embodiments, reads 112-1, 112-2, 112-n may be in a compressed format.

[0080] The secondary analysis unit 130 may receive the reads 112-1, 112-2, 112-n and store the reads 112-1, 112-2, 122-n in the memory 132 of the secondary analysis unit 130. In some implementations, this may include the sequencer 110 control program transferring the reads 112-1, 112-2, 112-n to the memory 132 of the secondary analysis unit 130. In other implementations, the secondary analysis unit 130 may request reads 112-1, 112-2, 122-n. In the case of compressed reads 112-1, 112-2, 112-n, the programmable logic device 134 of the secondary analysis unit 130 may be placed in a state B to act as decompressor 138 and may be used to decompress reads 112-1, 112-2, 112-n. The programmable logic device 134 may then be reconfigured in state A to act as a read aligner and used to align reads 112-1, 112-2, 112-n to the reference sequence.

[0081] The secondary analysis block 130 may be returned to state B to operate as a compression block, and such a compression block may be used to compress the aligned reads to prepare the aligned reads for transmission to computer 320. In this example, compression of the first packet of aligned reads includes compression of non-aligned reads. only the aligned reads, but also the data prepared by the read alignment unit 136 related to the aligned reads that will be used for gene fusion analysis. Such data is described with reference to the system 100 shown in FIG. 1, and may include, for example, the frequency of variant alleles, the number of unique read alignments, transcript read coverage, MAPQ score, data indicating homology between parental genes, or a combination thereof. In addition, other data that may be compressed into the aligned reads of the first batch may include (i) the candidate fusion reads, (ii) the region to which the potential fusion reads were aligned, and (iii) annotations of the reference genome segments to which the particular candidate was aligned. gene fusion. In some embodiments, the annotations may include annotations of gene exons, annotations indicating the presence of homologous genes, annotations indicating a list of enriched genes, or a combination thereof.

[0082] After compressing the aligned reads, secondary analysis unit 130 may store a first batch of compressed reads in memory 120. Sequencer 110 may then transmit the first batch 125 of aligned reads to computer 320 over network 310 for gene fusion analysis. Network 310 may include one or more wired networks, one or more wireless networks, or a combination thereof. In various embodiments, network 310 may be one or more of a wired Ethernet network, a wired optical network, a LAN, a WAN, a cellular network, the Internet, or a combination thereof. In some embodiments, computer 320 may be a remote cloud server. However, in other embodiments, computer 320 may communicate with sequencer 110 through a direct connection, such as a direct Ethernet connection, a USB-C connection, or the like. Although in this example in FIG. 300 compresses the first packet of reads before transmission; compression is optional. In contrast, compression is used as a way to reduce network bandwidth usage and storage costs, which can provide significant technological benefits and reduce costs when working with genomes with large data sizes.

[0083] In some embodiments, the first batch of aligned reads includes the entire set of reads produced for sample 105. In other embodiments, the first batch of aligned reads represents only a portion of the complete set of reads generated for sample 105, and a system can be used to facilitate parallel processing batch processing. For example, in some implementations, after the secondary analysis unit stores the first batch of aligned reads in memory 120, the secondary analysis unit 130 receives a second batch of reads that have not yet been aligned for storage in memory 132. Then, secondary analysis unit 130 The analysis may perform decompression if the second batch of reads has been compressed and alignment of the second batch of reads while computer 320 performs gene fusion analysis on the first batch of reads. This parallel processing, which is provided by batch processing of reads, can significantly reduce the run time of the system 300 required to identify confirmed gene fusions for probe 105 reads.

[0084] Computer 320 may receive the first batch of reads 125 over network 310 and store the first batch of reads in memory 320. In the case of a compressed first batch of reads 125, computer 320 may use a compress/decompress module 325 to decompress the first batch of reads and store the first batch of reads. in storage device 320. Computer 320 may then orchestrate an in-line gene fusion analysis process using a potential fusion identification module 140, a potential fusion filtering module 150, a feature set generation module 160, a machine learning model 170, a gene fusion detection module 180, and an output API module 190. in the same manner as described with reference to system 100 in FIG. 1.

[0085] The output data 192 may be transmitted to a variety of different devices via the network 310. As an example, the output data may be transmitted to a sequencer for display on the sequencer display 195. In an alternative or additional embodiment, output 192 may be output for display on the display of user device 330 via network 310. User device 330 may include a smartphone, tablet computer, laptop, desktop computer, or any other computer with a display. In an alternative or additional embodiment, output 192 may also be transmitted for output to printer 340 over network 310. In such embodiments, the output may be a printed copy of a report of identified confirmed gene fusions.

[0086] In FIG. 4 shows a block diagram of system components that can be used to implement a rapid gene fusion detection system.

[0087] The computing device 400 should be various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blades, mainframes, and other suitable computers. The computing device 450 should be various forms of mobile devices, such as personal digital assistants, cell phones, smartphones, and other similar computing devices. Additionally, computing device 400 or 450 may include universal serial bus (USB) flash drives. USB flash drives can store operating systems and other applications. USB flash drives may include I/O components, such as a wireless transmitter or a USB connector that can be inserted into a USB port on another computing device. The components, their connections and relationships, and their functions presented herein are examples only and are not intended to limit the embodiments described and/or claimed herein.

[0088] Computing device 400 includes a processor 402, memory 404, storage device 406, a high-speed interface 408 coupled to memory 404 and high-speed expansion ports 410, and a low-speed interface 412 coupled to low-speed bus 414 and storage device 408. Components 402, 404, 406, 408, 410 and 412 are all connected to each other using various buses, and can be mounted on a common motherboard or in other ways if necessary. Processor 402 may process instructions intended to be executed on computing device 400, including instructions stored on memory device 404 or storage device 408, to display graphical user interface information on an external input/output device, such as display 416 coupled to a high-speed interface 408. In other embodiments, multiple processors and/or multiple buses may be used, along with multiple storage devices and memory types, as needed. In addition, multiple computing devices 400 may be connected, with each device providing a site of required operations, such as a bank of servers, a bank of blade servers, or a multiprocessor system.

[0089] Information in memory 404 is stored in computing device 400. In one embodiment, memory 404 is a block or blocks of volatile memory. In another embodiment, memory 404 is a block or blocks of non-volatile memory. Memory 404 may also be another form of computer-readable media, such as a magnetic or optical disk.

[0090] The storage device 408 is configured to provide a large storage capacity for the computing device 400. In one embodiment, the storage device 408 may be or comprise a computer-readable medium such as a floppy disk device, a hard disk device, an optical disk device, or a tape device, flash memory, or other similar solid-state storage device or array of devices, including devices in a storage area network or other configurations. The computer program product may be practically implemented on a storage medium. The computer program product may also contain instructions that, when executed, perform one or more methods such as those described above. The storage medium is a computer-readable or computer-readable medium, such as memory 404, storage device 408, or memory on processor 402.

[0091] A high-speed controller 408 manages high-bandwidth operations for computing device 400, and a low-speed controller 412 manages lower-bandwidth operations. This distribution of functions is just an example. In one embodiment, the high-speed controller 408 is coupled to memory 404, display 416, such as through a graphics processor or accelerator, and high-speed expansion ports 410 into which various expansion cards (not shown) can be installed. In an embodiment, low-speed controller 412 is coupled to storage device 408 and low-speed expansion port 414. The low-speed expansion port, which may include various communications ports, such as USB, Bluetooth, Ethernet, wireless Ethernet, may be connected to one or more input/output devices, such as a keyboard, pointing device, microphone/speaker pair, scanner, or a network device such as a switch or router, such as a network adapter. Computing device 400 may be implemented in several different forms, as shown in the figure. For example, it may be implemented as a standard server 420 or multiple times as a group of such servers. It may also be implemented as part of a rack server system 424. Additionally, it may be implemented on a personal computer such as a laptop 422. In an alternative embodiment, components of the computing device 400 may be combined with other components in a mobile device (not shown) such as device 450. Each of such devices may contain one or more computing devices 400, 450, and the entire system may consist of a plurality of computing devices 400, 450 communicating with each other.

[0092] Computing device 400 may be implemented in several different forms, as shown in the figure. For example, it may be implemented as a standard server 420 or multiple times as a group of such servers. It may also be implemented as part of a rack server system 424. Additionally, it may be implemented on a personal computer such as a laptop 422. In an alternative embodiment, components of the computing device 400 may be combined with other components in a mobile device (not shown) such as device 450. Each of such devices may contain one or more computing devices 400, 450, and the entire system may consist of a plurality of computing devices 400, 450 communicating with each other.

[0093] The computing device 450 includes a processor 452, a memory 464, and an input/output device such as, but not limited to, a display 454, a communications interface 466, and a transceiver 468. The device 450 may also be provided with a storage device such as a microdrive or other device to provide additional storage capacity. The components 450, 452, 464, 454, 466, and 468 are all connected to each other using various buses, and some of the components may be mounted on a common motherboard or in other ways, as appropriate.

[0094] The processor 452 may execute instructions contained in the computing device 450, including instructions stored in the memory device 464. The processor may be implemented as a chipset, including individual and multi-component analog and digital processors. In addition, the processor may be implemented using any of a number of architectures. For example, processor 410 may be a CISC (full instruction set computing) processor, a RISC (reduced instruction set computing) processor, or a MISC (minimum instruction set computing) processor. The processor may provide, for example, coordination of other components of device 450, such as managing user interfaces, running applications by device 450, and communicating wirelessly with device 450.

[0095] The processor 452 may communicate with the user through a control interface 458 and a display interface 456 coupled to the display 454. The display 454 may be, for example, a TFT (thin film transistor) type display or an OLED (organic light emitting diode) type display. ), or other suitable display technologies. The display interface 456 may include suitable circuitry for driving the display 454 and for presenting graphical and other information to the user. The control interface 458 may receive commands from the user and convert them for transmission to the processor 452. Additionally, an external interface 462 may be provided in communication with the processor 452 to provide the device 450 with near-field communications with other devices. In some embodiments, external interface 462 may provide, for example, wired communications, or in other embodiments, wireless communications, and a variety of interfaces may also be used.

[0096] Information in storage device 464 is stored in computing device 450. Storage device 464 may be implemented as one or more computer-readable media or media, a block or blocks of volatile memory, or a block or blocks of non-volatile memory. Additionally, expansion memory 474 may be provided coupled to device 450 via expansion interface 472, which may include, for example, a SIMM (single inline memory module) card interface. Such expansion memory 474 may provide device 450 with additional storage space or may also store applications or other information for device 450. In particular, expansion memory 474 may include instructions to perform or supplement the processes described above, and may also include secure information. Thus, for example, expansion memory 474 may be provided as a secure module for device 450 and may be programmed with instructions that allow device 450 to be used in a secure manner. In addition, SIMM cards can transfer secure applications as well as additional information, such as placing identity information on the SIMM card without the possibility of compromise.

[0097] The memory may include, for example, flash memory and/or NVRAM memory, as described below. In one embodiment, the computer program product is practically implemented on a storage medium. The computer program product contains instructions that, when executed, perform one or more methods such as those described above. The information medium is a computer-readable or machine-readable medium, such as storage device 464, expansion memory 474, or memory on processor 452, where the information may be received, for example, by transceiver 468 or external interface 462.

[0098] Device 450 may communicate wirelessly through a communications interface 466, which may include circuitry for digital signal processing if desired. The communications communications interface 466 may provide communications in various modes or protocols, such as, but not limited to, GSM voice calls, SMS, EMS or MMS messages, CDMA, TDMA, PDC, WCDMA, CDMA2000, or GPRS. Such data transmission can be accomplished, for example, using a radio frequency transceiver 468. In addition, short-range communication is possible, for example, using Bluetooth, Wi-Fi, or another similar transceiver (not shown). In addition, GPS (Global Positioning System) receiver module 470 may transmit additional navigation and location-related wireless data to device 450 that can be used as needed by applications running on device 450.

[0099] Device 450 may also transmit an audio signal using an audio codec 460, which can receive speech information from a user and convert it into usable digital information. Audio codec 460 may also generate an audio signal for the user, for example, through a speaker, such as through the headset of device 450. Such audio signal may be audio from voice telephone calls, may be recorded audio, such as voice messages, music files, and the like. , and may also include audio generated by applications running on device 450.

[0100] Computing device 450 may be implemented in several different forms, as shown in the figure. For example, it may be implemented as a cellular telephone 480. It may also be implemented as part of a smartphone 482, a personal digital assistant, or other similar mobile device.

[0101] Various implementations of the systems and methods described herein may be implemented in the form of a digital electronic circuit, an integrated circuit, custom ASICs (application integrated circuits), computer hardware, firmware, software, and/or as a combination of such implementation options. These various implementations may include implementation in one or more computer programs configured to be executed and/or interpreted on a programmable system including at least one programmable processor, which may be specialized or general purpose, coupled to a storage system to receive data from it. and commands and transmitting data and commands thereto, at least one input device and at least one output device.

[0102] These computer programs (also referred to as programs, software, software applications or code) include machine instructions for a programmable processor and may be implemented in a high-level procedural and/or object-oriented programming language, and/or assembly/machine language . As used herein, the terms “machine-readable medium”, “computer-readable medium” refer to any computer program product, apparatus and/or device, such as magnetic disks, optical disks, storage media, programmable logic devices (PLDs), used for transmitting computer data. instructions and/or data to a programmable processor, including a computer-readable medium, that receives the machine instructions as a computer-readable signal. The term "machine-readable signal" refers to any signal used to convey machine instructions and/or data to a programmable processor.

[0103] To provide user interaction, the systems and methods described herein may be implemented on a computer having a display device, such as a CRT (cathode ray tube) or LCD (liquid crystal) monitor, to display information to the user, and a keyboard and A pointing device, such as a mouse or trackball, that allows the user to provide input to the computer. Other types of devices can also be used to provide user interaction; the feedback provided to the user can be in any tangible form, such as visual feedback, auditory feedback or haptic feedback; and input from the user may come in any form, including audio, speech, or tactile input.

[0104] The systems and methods described herein may be implemented in a computer system that includes a server component, such as a data server, or includes an intermediate component, such as an application server, or includes a client component (for example, a client computer with a graphical user interface or web browser through which a user can interact with an implementation of the systems and methods described herein), or any combination of such server, middleware, and client components. System components can be interconnected by any form or medium of digital data exchange, such as a communications network. Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0105] The computer system may include clients and servers. The client and server are essentially remote from each other and typically communicate through a communications network. The functional dependence of the client and server arises due to computer programs running on the respective computers and having a client-server functional dependence on each other.

OTHER IMPLEMENTATION OPTIONS

[0106] A number of embodiments are described. However, it will be appreciated that various modifications are possible without departing from the spirit and scope of the invention. Additionally, the logical flows depicted in the figures do not require the specific or sequential order shown to achieve the desired results. In addition, other steps may be provided or omitted from the described flows, and other components may be added to or removed from the described systems. Accordingly, the following claims cover other embodiments.

Claims (44)

1. A computer-implemented method for identifying one or more gene fusions in a biological sample (105), comprising: for each reference sequence position from a plurality of reference sequence positions: obtaining first data, which represents an accumulation of aligned reads obtained by short read sequencing and having a high depth of coverage at a specified position of the reference sequence, from the read alignment block, by one or more computers; identifying multiple potential gene fusions included in the first data obtained by one or more computers (140); filtering a plurality of potential gene fusions by one or more computers (150) to remove duplicate potential fusions that result from high depth of coverage at a specified reference sequence position, thereby determining a filtered set of potential gene fusions; for each specific potential gene fusion from the filtered set of potential gene fusions: generating input data (162) by one or more computers (160) for input to a machine learning model (170), wherein generating the input data includes retrieving characteristic data distinguishing a particular candidate gene fusion from data including: (i) one or more reference sequence segments to which the particular candidate gene fusion was aligned using the read alignment block (136), and (ii) data generated based on the output of the read alignment block (136), wherein the data generated based on the output of the read alignment block includes one or more of an allele variant frequency, a number of unique read alignments, a MAPQ score, or data, indicating homology between parental genes; transmission by one or more computers of generated input data (162) as input to a machine learning model (170), wherein the machine learning model is trained using labeled training vectors, where each labeled training vector represents a training fusion candidate and contains (i) one or more reference sequence segments to which the training candidate gene fusion was aligned, and (ii) one or more of the number of variant allele frequencies, the number of unique read alignments, a MAPQ score, or data indicating homology between parental genes, generate output data (178 ), reflecting the probability that a potential gene fusion is a confirmed gene fusion based on the results of applying a machine learning model (170) to the input data (162) representing (i) one or more reference sequence segments to which a particular a potential gene fusion using the read alignment block (136), and (ii) data generated based on the output of the read alignment block (136), wherein the data generated based on the output of the read alignment block includes any one or more of a frequency occurrence of variant alleles, number of unique read alignments, MAPQ score or data indicating homology between parental genes; receiving by one or more computers (180) output data (178) generated by the machine learning model (170) after processing the generated input data (162) by the machine learning model; And determining whether a particular candidate fusion is a confirmed gene fusion based on the output (178) by one or more computers (180). 2. Method according to claim 1, wherein generating the input data (162) further includes retrieving feature data, which includes annotation data describing the annotations of the reference sequence segments to which the particular candidate gene fusion was aligned using the read alignment block (136); And wherein the machine learning model (170) is trained to generate output (178) reflecting the probability that the candidate gene fusion is a confirmed candidate gene fusion based on the results of applying the machine learning model (170) to the input data (162) being: (i) one or more reference sequence segments to which the particular candidate gene fusion was aligned using the read alignment block (136), (ii) annotation data describing the annotations of the reference sequence segments to which the particular candidate gene fusion was aligned using the read alignment block (136), and (iii) data generated based on the output of the read alignment unit (136). 3. The method of any one of the preceding claims, wherein identifying a plurality of potential gene fusions included in the acquired first data by one or more computers (140) includes identifying a plurality of split read alignments by one or more computers. 4. The method of any one of the preceding claims, wherein identifying a plurality of potential gene fusions included in the acquired first data by one or more computers (140) includes identifying a plurality of alignments of discordant read pairs by one or more computers. 5. The method as claimed in any one of the preceding claims, wherein the read aligner (136) is implemented using a set of one or more processing modules that are configured to use hardware logic circuits physically arranged to perform operations using the hardware logic circuits, to : (i) receiving data representing the first reading, (ii) matching the data representing the first read to one or more regions of the reference sequence to identify one or more matching positions of the reference sequence, (iii) generating one or more alignment estimates corresponding to each of the matching positions of the reference sequence for the first read, (iv) selecting one or more potential alignments for the first read based on the one or more alignment scores, and (v) outputting data representing a candidate alignment for the first read. 6. Method according to any one of paragraphs. 1-4, in which the read alignment unit (136) is implemented using a set of one or more processing modules through the use of one or more central processors, i.e. CPU, or one or more graphics processors, i.e. GPU, to execute software instructions that cause one or more CPUs or one or more GPUs to execute functions: (i) receiving data representing the first reading, (ii) matching the data representing the first read to one or more reference sequence regions to identify one or more matching reference sequence positions for the first read, (iii) generating one or more alignment estimates corresponding to each of the matching positions of the reference sequence for the first read, (iv) selecting one or more potential alignments for the first read based on the one or more alignment scores, and (v) outputting data representing a candidate alignment for the first read. 7. The method as claimed in any one of the preceding claims, wherein obtaining by one or more computers first data that represents an accumulation of aligned reads from a read alignment unit (136) comprises obtaining by one or more computers an accumulation of aligned reads from a storage device (120) and performing one or more operations according to claim 1, while the read alignment unit (136) aligns the second accumulation of reads that have not yet been aligned. 8. The method of any one of the preceding claims, wherein the data generated based on the output of the read alignment block includes any one or more of a number of variant allele frequencies, a number of unique read alignments, a MAPQ score, or data indicating homology between parent genes . 9. The method of any one of the preceding claims, wherein determining whether a particular candidate fusion matches a confirmed candidate gene fusion based on the output data (178) comprises: determining by one or more computers (180) whether the output data (178) satisfies a predetermined threshold value; And determining that a particular candidate fusion matches a confirmed candidate gene fusion based on determining that the output (178) satisfies predefined thresholds. 10. The method of claim 1, wherein determining whether a particular candidate fusion matches a confirmed gene fusion based on the output data (178) includes: determining by one or more computers (180) whether the output data (178) satisfies a predetermined threshold value; And determining that a particular candidate fusion does not correspond to a confirmed candidate gene fusion based on determining that the output (178) does not meet predefined thresholds. 11. The method of claim 1, wherein the high depth of coverage at the specified position of the reference sequence is equal to 30x coverage. 12. A system (100) for identifying one or more gene fusions in a biological sample, comprising: one or more computers and one or more storage devices on which instructions are stored, when executed by one or more computers, one or more computers perform operations, including the method of any one of claims. 1-11. 13. A non-transitory computer-readable medium for identifying one or more gene fusions in a biological sample, on which is stored software containing instructions executable by one or more computers, by executing which one or more computers perform operations, including a method according to any one of pp. 1-11.