nearproteins

Store for finding similar amino acid sequences using Locality Sensitive Hashing and Approximate Nearest Neighbors.

Installation

Python, with dependencies:

• Annoy
• BioPython

pip install -r requirements.txt

Redis

Instructions

1. Load proteins into database

$ ./load-proteins < data/proteins.fasta

2. Query database

$ ./query-proteins < data/proteins.fasta # returns JSON for each record

Python API

Very basic. I plan to add more configuration.

Loading data into store

from Bio import SeqIO
import nearproteins

store = nearproteins.SimilarStringStore()

store.engine.clean_all_buckets()

records = SeqIO.parse(handle, 'fasta')

for record in records:
    store.add(str(record.seq), record.id)

Retrieving records from store

# returns array of vectors, match IDs, similarities
results = store.get(str(record.seq)) 

Use as a server

You can query and add records to the database using simple sockets.

$ ./server # start the server, listens on port 1234

In another window...

$ nc 127.0.0.1 1234 # connect
SET 1 AUSTIN
SET 2 BOSTON
GET AUSTIN
{"1": 0.0}
GET BOSTON
{"2": 0.0}

You can use this to build a simple client in another language such as Ruby

require 'dna'
require 'socket'
require 'json'

HOSTNAME = '127.0.0.1'
PORT = '1234'

socket = TCPSocket.open HOSTNAME, PORT

File.open('proteins.fasta') do |handle|
  records = Dna.new handle, :format => :fasta

  records.each do |record|
    socket.puts "GET #{record.sequence}"
    resp = JSON.parse(socket.gets)
    p resp
  end
end