Systems and methods for peptide identification

A method for rapid and efficient peptide identification from large-scale mass spectrometry data through high performance computing database peptide search, the method comprising:

• providing a symmetric multiprocessor supercomputer comprising a plurality of processors and shared resources, the shared resources comprising a shared memory storage in operable communication with the plurality of processors, and at least one machine-readable medium in operable communication with the plurality of processors;
• and
• performing the following supersteps:
• (a) extracting, by at least one processor of the plurality of processors, a plurality of peptide database partitions (PDBs) from a peptide database and then indexing the plurality of PDBs, the plurality of PDBs being extracted by clustering data entries of the peptide database based at least on a Mod Distance (Δm) metric and load balancing the clustered data entries evenly across the PDBs;
• (b) providing, by at least one processor of the plurality of processors, a plurality of experimental spectra, each spectrum of the plurality of experimental spectra comprising a batch of pre-processed mass spectrometry data from a mass spectrometry data set of a database;
• (c) completing, by at least one processor of the plurality of processors, a peptide search of the plurality of spectra against the plurality of PDBs to produce a plurality of partial results and serializing the plurality of partial results;
• and
• (d) assembling, de-serializing, and synchronizing, by at least one processor of the plurality of processors, the plurality of partial results to form a complete result for identifying peptides, and writing the complete result to a file system, thus providing the rapid and efficient peptide identification from large-scale mass spectrometry data,
• the peptide search further comprising the following steps:
• (i) loading the plurality of spectra into a plurality of forward queues (qf) in the shared memory storage;
• (ii) reading the spectra from each respective forward queue in the plurality of forward queues, recycling the forward queue as a return queue (qr), searching the spectra against the plurality of PDBs to produce the plurality of partial results, and writing the plurality of partial results to a plurality of intermediate queues (qk) in the shared memory storage;
• and
• (iii) reading the partial results from each respective intermediate queue in the plurality of intermediate queues, recycling each intermediate queue as an intermediate return queue (qk′), serializing the plurality of partial results, and writing the plurality of partial results to the shared memory storage.

A system for rapid and efficient peptide identification from large-scale mass spectrometry data through high performance computing database peptide search, the system comprising:

• a symmetric multiprocessor supercomputer comprising a plurality of processors and shared resources, the shared resources comprising a shared memory storage in operable communication with the plurality of processors;
• and
• at least one machine-readable medium in operable communication with the plurality of processors, the at least one machine-readable medium having instructions stored thereon that, when executed by one or more processors of the plurality of processors, perform the following supersteps:
• (a) providing, by the one or more processors, a plurality of peptide database partitions (PDBs) from a peptide database and then indexing the plurality of PDBs, the plurality of PDBs being extracted by clustering data entries of the peptide database based at least on a Mod Distance (Δm) metric and load balancing the clustered data entries evenly across the PDBs;
• (b) providing, by the one or mare processors, a plurality of experimental spectra, each spectrum of the plurality of experimental spectra comprising a batch of pre-processed mass spectrometry data from a mass spectrometry data set of a database;
• (c) completing a peptide search of the plurality of experimental spectra against the plurality of PDBs to produce a plurality of partial results and serializing the plurality of partial results;
• and
• (d) assembling, de-serializing, and synchronizing the plurality of partial results to form a complete result for identifying peptides, and writing the complete result to a file system, thus providing the rapid and efficient peptide identification from large-scale mass spectrometry data;
• the superstep (a) comprising peptide database clustering based on the Mod Distance (λm), given as:
• Δ
• ⁢
• ⁢
• m
• ⁡
• (
• x
• ,
• y
• )
• =
• 2
• -
• a
• max
• (
• len
• ⁡
• (
• x
• )
• ,
• len
• ⁡
• (
• y
• )
• for a pair of peptide database entries (x, y), where the sum of unedited letters from both sequence termini is (a),
• each of the plurality of processors having a main memory locally connected, and each of the plurality of experimental spectra provided in superstep (b) having a size selected to fit within the main memory on one of the plurality of processors,
• the partial database peptide search comprising the following steps:
• (i) loading the plurality of experimental spectra into a plurality of forward queues (qf) in the shared memory storage;
• (ii) reading the plurality of experimental spectra from each respective forward queue in the plurality of forward queues, recycling the forward queue as a return queue (qr), searching the spectra against the plurality of PDBs to produce the plurality of partial results, and writing the plurality of partial results to a plurality of intermediate queues (qk) in the shared memory storage;
• and
• (iii) reading the partial results from each respective intermediate queue in the plurality of intermediate queues, recycling each intermediate queue as an intermediate return queue (qk′), serializing the plurality of partial results and writing the plurality of partial results to the shared memory storage, and
• the superstep (c) comprising actively allocating shared resources across steps (i), (ii), and (iii) to optimize performance in a multi-threaded, multi-core, shared memory, parallel process, producer-consumer pipeline.