Bioinformatics systems, apparatuses, and methods executed on an integrated circuit processing platform
Inventors
Van Rooyen, Pieter • McMillen, Robert J. • Ruehle, Michael
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-10262105-B2
Publication Date
2019-04-16
Expiration Date
Abstract
A system, method and apparatus for executing a sequence analysis pipeline on genetic sequence data includes a structured ASIC formed of a set of hardwired digital logic circuits that are interconnected by physical electrical interconnects. One of the physical electrical interconnects forms an input to the structured ASIC connected with an electronic data source for receiving reads of genomic data. The hardwired digital logic circuits are arranged as a set of processing engines, each processing engine being formed of a subset of the hardwired digital logic circuits to perform one or more steps in the sequence analysis pipeline on the reads of genomic data. Each subset of the hardwired digital logic circuits is formed in a wired configuration to perform the one or more steps in the sequence analysis pipeline.
Core Innovation
The invention relates to an apparatus and genomics analysis platform for executing one or more steps of a sequence analysis pipeline on genetic sequence data of a subject. The genetic sequence data comprises one or more reads of genomic data, one or more genetic reference sequences, and an index of the genetic reference sequences. The system includes an integrated circuit comprising a set of reconfigurable hardwired digital logic circuits interconnected by physical electrical interconnects, where the hardwired digital logic circuits operate in a wired configuration to receive location information and access reference segments and align reads to the reference segments.
For alignment, the invention generates a linear array of scoring cells to calculate alignment scores within a band of a two-dimensional alignment matrix. The two-dimensional alignment matrix has a first dimension representing a query segment from the read and a second dimension representing the reference segment, and each scoring cell transmits scores to neighboring cells, receives scores from neighboring cells, and computes a new score using a match score based on a comparison of the query base and the reference base. Positions of the new scores computed by the scoring cells represent a wavefront traversing the two-dimensional alignment matrix.
The wavefront traversal is driven by sequentially shifting query bases through the linear array in a first direction and reference bases through the linear array in a second direction opposite the first direction, with each shift corresponding to a step of the wavefront along the corresponding dimension. The pipeline further includes mapping and alignment module stages, followed by variant calling using a Hidden Markov Model that compares candidate haplotype sequences to nucleotide sequences from reads and determines a probability to assess whether the candidate haplotype sequence is a true representation of the subject's genetic sequence data.
Claims Coverage
The independent claims are directed to wired reconfigurable hardwired digital logic alignment, mapping and alignment with HMM-based variant calling, and a cloud-accessible platform combining CPU/GPU with FPGA-based hardwired logic and shared memory. The inventive coverage centers on wavefront traversal in a banded two-dimensional alignment matrix and on Hidden Markov Model variant calling.
Wired reconfigurable hardwired digital logic alignment with wavefront traversal
An apparatus with an integrated circuit of reconfigurable hardwired digital logic circuits interconnected by physical electrical interconnects, operating in a wired configuration to receive location information specifying a reference segment of genetic reference sequences, access reference segments, and align genomic reads to the reference segments. The aligning generates a linear array of scoring cells to calculate alignment scores within a band of a two-dimensional alignment matrix using match scores based on comparison of query base and reference base, and the computed scoring positions represent a wavefront traversing the two-dimensional alignment matrix.
Directional wavefront traversal with opposite query/reference shifting
The aligning includes shifting query bases sequentially from the query sequence through the linear array in a first direction and shifting reference bases sequentially from the reference segment through the linear array in a second direction opposite the first direction, with each shift corresponding to a wavefront step along the respective dimension of the alignment matrix.
Hardwired processing engines for mapping, alignment, and HMM-based variant calling
An apparatus with an integrated circuit formed of hardwired digital logic circuits interconnected by physical electrical interconnects including a memory interface, arranged as processing engines. The processing engines include a mapping module to map a selected read to one or more segments of genetic reference sequences based on an index to produce a mapped read, an alignment module to align the selected mapped read to one or more positions in the segments to produce an aligned read, and a variant calling module that implements a Hidden Markov Model to compare a candidate haplotype sequence to a read's nucleotides and determine a probability for whether the candidate haplotype sequence is a true representation of the subject's genetic sequence data.
Cloud accessible CPU/GPU plus FPGA platform with shared memory executing HMM
A genomics analysis platform comprising a cloud accessible server with one or more integrated circuits forming a CPU or GPU responsive to software algorithms to perform a first set of genomic processing steps, and one or more FPGA integrated circuits configured by firmware to arrange hardwired digital logic circuits interconnected by physical interconnects to perform a second set of genomic processing steps. The second set includes implementing a Hidden Markov Model on genetic sequence data, and the platform includes memory accessible by each CPU or GPU and each FPGA via physical interconnects to store genetic sequence data and result data.
Across the independent claims, the core coverage centers on executing sequence analysis pipeline steps using hardwired digital logic and structured alignment computation with wavefront traversal over a two-dimensional alignment matrix, combined with mapping and alignment stages and variant calling via a Hidden Markov Model. The platform claim further specifies a cloud-accessible arrangement of CPU/GPU and firmware-configured FPGA hardwired logic with shared memory accessible to all processing units.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?