Depends on #4

Fetch interactions for network of interests

Implement the following methods.

• Get query for one organism and one protein

get-query-by-taxon-and-prot [taxid protId] 
Same as above but with restriction on interactor identifiers
Input: 9606 "P04040"
Output: [:and [:taxidA 9606] [:taxIdB 9606] [:species 9606] [:or [:idA "P04040"] [:idB "P04040"]]]

• Get queries for one taxon and multiple proteins

get-queries-by-taxon-and-prot-pool [taxId protPool limit]
A query should be less thatn 1000 characters so we have to split it into a list of queries.
The queries have the same taxid restrictions but different interactor identifier restrictions (in "or" section).
The following example aims to explain how to split but in this case we could have keep it in one query.
As an indicator, a query should stick to around 20 `and` blocks.
Input: 9606 ["P04040" "Q9D2V5"]
Output:
[[:and 
    [:taxidA 9606] 
    [:taxIdB 9606] 
    [:species 9606] 
    [:or 
        [:and [:idA "P04040"] [:idB "Q9D2V5"]]
        [:and [:idA "Q9D2V5"] [:idB "P04040"]]]]
[:and 
    [:taxidA 9606] 
    [:taxIdB 9606] 
    [:species 9606] 
    [:or 
        [:and [:idA "P04040"] [:idB "P04040"]]
        [:and [:idA "Q9D2V5"] [:idB "Q9D2V5"]]]]]

• Format query in MIQL format

Update to-miql [query]

• Fetch interactions by query

Update fetch-by-query [client query]

• Import ppi-query as a maven module inside PMB
• Test invoking ppi-query function from PMB
• Define interface between ppi-query and PMB
  • Data interface for proteins, interactions, etc. (use Map? use Java interface?)
  • Program interface for processes (Java interface for fetching networks mostly)
• Deeper integration
  • Integrated progress report from ppi-query to PMB

For now the cache is in memory and grows indefinitely
TODO

• Add cache persistance (to disk)
• Flatten cache for easier management
  {:org {:prot :ortho-group}} => {[:org :prot] :ortho-group}
• Set up caching strategy
• Link memory cache, disk cache and inparanoid webservices all together

Pseudo-code:

(if in-memory-cache?
   (get-from-memory-cache)
   (save-in-memory-cache
     (if in-disk-cache?
        (get-from-disk-cache)
        (save-in-disk-cache
          (get-from-inparanoid-ws)))))

Proposed caching strategy:

• LRU with size threshold (with clojure core.cache)
• Write to memory and disk on new entry (or periodically write to disk?)

Persistence:

• Use nippy for writting/reading clojure data structures in binary
• Split the cache in organism pairs (like in PMB)

TODO cache loader:

• Download OrthologXML by organism pairs
• Parse & load to cache
• Add approx loaded percent
• Add progress monitoring interface

• Fetch registry xml
• Parse to maps
• Store & update memory cache
• Fetch from PMB
• disk cache? (like ortholog cache)

Fetch interactions for interactome

Implement the following methods.

• Get query by taxon

get-query-by-taxon [taxId]
Returns query filtering interaction for which the taxonomic ID of the two proteins and
 the species is the `taxId` parameter
Input: 9606
Output: [:and [:taxidA 9606] [:taxidB 9606] [:species 9606] ]

• Format query in MIQL format

to-miql [query]
Returns the query in MIQL format
Example: [:and [:key1 "value"] [:or [:key2 "value2"] [:key3 "value3"]] ] returns "key1:value and (key2:value2 or key3:value3)"

• Fetch interactions by query

fetch-by-query [client query]
Returns result of a query (handling pagination if needed)