X-Y-C.rkt

;;; X-Y-C queries
;;; unknown-unknown-known concepts
;;;
;;; example:
;;;
;;; X - increases - G1 - increases - G2
;;;
;;; where G1 and G2 are genes.
;;;
;;; Also want to report what other things that G1 increases (as part
;;; of tox info)


;;; Hmm--need to be smarter here!
;;;
;;; Suppose we want to increase expression of gene G2, and that genes
;;; A, B, and C increase expression of G2, and genes D, E, and F
;;; decrease expression of G2.  It may be that a drug D increases
;;; expression of A, but decreases expression of B and C, and
;;; increases expression of D, E, and F!  Since we don't have
;;; quantitative data, we have to assume this would be
;;; counter-productive!
;;;
;;; So, really, we need to calculate the effects of D on *all* genes
;;; known to regulate G2, use this information for ranking, and
;;; display the info as part of the output.  And, of course, we need
;;; to consider the strength of the evidence for all of these claims.

;;; Also, the approach I'm taking with
;;; G1-G2-edges-without-duplicate-dbname/scui seems wrong.  Really
;;; need to separate evidence of G1 increasing G2 from evidence that D
;;; increases G1.  After all, the evidence that G1 increases G2 can
;;; already be complex: should be calculating all the ways in which G1
;;; positively or negatively affects G2, weighing the evidence, etc.,
;;; and producing a full mini-report just for the 'G1 increases G2'
;;; claim.  Then the user can pick a G2, and get a full mini-report
;;; for each D that interacts with G2.  This is probably easier in a
;;; GUI or custom text interface than in the REPL.

;;; And still need to create the 'tox report' showing what other genes
;;; G1 affects, what other genes D affects, etc.  Probably also want
;;; to list what diseases/symptoms both G1 and D are associated with,
;;; ranked by evidence.

;;; Also, given how noisy semmed is, prolly want to cross-reference
;;; the genes against the NLM list of genes/pubmed ids to flag/filter
;;; those associations that are bogus, picked up through bad NLP.
;;; May be best to incorporate this list as an actual relation we
;;; can query over, although we might need to index it as well.
;;; This would be the most consistent and flexible approach, though,
;;; as opposed to putting the info in a hash table or whatever.

;;; Example problems:
;;;
;;; X that downregulates GMPPB resulting in a decrease of GMPPA expression
;;; X that increases KLF4 in order to compensate for a decrease in SOX5.
;;; X that increases the expression of the CDKL5 gene to increase CDKL5 protein expression

#lang racket

(provide
  (all-from-out "mk.rkt")
  (all-from-out "mk-db.rkt")
  (all-from-out "common.rkt")
  (all-defined-out))

(require
  "mk.rkt"
  "db.rkt"
  "mk-db.rkt"
  "common.rkt"
  (except-in racket/match ==)
  (only-in srfi/1 iota))

(define X-Y-C_VERSION_STRING "X-Y-C Tool 0.1.0")

(define argv (current-command-line-arguments))
(define argv-optional '#(CONFIG_FILE))

(when (not (<= (vector-length argv) (vector-length argv-optional)))
  (error "optional arguments ~s; given ~s" argv-optional argv))

(displayln "Starting X-Y-C tool...")
(displayln X-Y-C_VERSION_STRING)
;; Loading will occur at first use if not explicitly forced like this.
(load-config #t (and (<= 1 (vector-length argv)) (vector-ref argv 0)))
(load-databases #t)

;;; Decreases/increases predicate names
(define DECREASES_PREDICATE_NAMES (config-ref 'decreases-predicate-names))
(define INCREASES_PREDICATE_NAMES (config-ref 'increases-predicate-names))



(define G-G-increases-predicate-names
  '("positively_regulates"
    ;; "causes"
    ;; "produces"  ;;; maybe??
    ;; "causes_condition"
    ;; "causally_related_to"
    ;; "contributes_to"
    ;; "causes_adverse_event"
    ;; "gene_associated_with_condition"
    ;; "gene_mutations_contribute_to"
    ;; "disease_to_gene_association"
    "increases_activity_of"
    "increases_expression_of"
    "increases_molecular_interaction"
    "increases_response_to"
    "increases_secretion_of"
    "increases_stability_of"
    "increases_synthesis_of"
    "increases_transport_of"
    "increases_uptake_of"
    "decreases_degradation_of"
    "posetively_regulates" ;;; robokop typo??
    ;; "positively_regulates__entity_to_entity"
    ;; "increases_molecular_modification_of"
    ;; "increases_localization_of"
    ;; "increases_splicing_of"
    ;; "decreases_mutation_rate_of"
    ;; "predisposes"
    ;; "decreases_metabolic_processing_of"
    ))

(define D-G-increases-predicate-names
  '("positively_regulates"
    ;; "causes"
    ;; "produces"  ;;; maybe??
    ;; "causes_condition"
    ;; "causally_related_to"
    ;; "contributes_to"
    ;; "causes_adverse_event"
    ;; "gene_associated_with_condition"
    ;; "gene_mutations_contribute_to"
    ;; "disease_to_gene_association"
    "increases_activity_of"
    "increases_expression_of"
    "increases_molecular_interaction"
    "increases_response_to"
    "increases_secretion_of"
    "increases_stability_of"
    "increases_synthesis_of"
    "increases_transport_of"
    "increases_uptake_of"
    "decreases_degradation_of"
    "posetively_regulates" ;;; robokop typo??
    ;; "positively_regulates__entity_to_entity"
    ;; "increases_molecular_modification_of"
    ;; "increases_localization_of"
    ;; "increases_splicing_of"
    ;; "decreases_mutation_rate_of"
    ;; "predisposes"
    ;; "decreases_metabolic_processing_of"
    ))

;;; X - increases - G1 - increases - G2



#|
;;; let's try NGLY1 as an example
(define G2-concepts
  '((semmed 141262 "UMLS:C1425023" "NGLY1 gene" (4 . "gene") (("umls_type_label" . "['Gene or Genome']") ("xrefs" . "['HGNC:HGNC:17646', 'OMIM:610661', 'MTH:NOCODE']") ("id" . "UMLS:C1425023") ("umls_type" . "['T028']") ("labels" . "['gene']")))
 (robokop 22660 "HGNC:17646" "NGLY1" (0 . "(\"named_thing\" \"gene\")") (("locus_group" . "protein-coding gene") ("chromosome" . "3") ("location" . "3p24.2") ("taxon" . "9606") ("id" . "HGNC:17646") ("equivalent_identifiers" . "(\"UniProtKB:H0Y2P2\" \"ENSEMBL:ENSG00000151092\" \"UniProtKB:Q96IV0\" \"UniProtKB:C9JU75\" \"HGNC:17646\" \"NCBIGENE:55768\" \"UniProtKB:A0A0C4DFP4\")")))
 (orange 16958 "NCBIGene:55768" "NGLY1" (6 . "(\"gene\")") (("iri" . "http://www.ncbi.nlm.nih.gov/gene/55768") ("synonym" . "(\"FLJ11005\" \"PNG1\" \"peptide-N(4)-(N-acetyl-beta-glucosaminyl)asparagine amidase\" \"Peptide-N-Glycanase 1, S. Cerevisiae, Homolog of\" \"N-GLYCANASE 1; NGLY1\" \"NGLY1\")") ("in_taxon" . "NCBITaxon:9606") ("description" . "N-glycanase 1") ("same_as" . "(\"ENSEMBL:ENSG00000151092\" \"HGNC:17646\" \"OMIM:610661\" \"Orphanet:406885\")") ("provided_by" . "(\"orphanet.ttl\" \"omim.ttl\")") ("id" . "NCBIGene:55768")))))
|#


;;; let's try BRCA1 as an example
(define G2-concepts
  '((semmed 74686 "UMLS:C0376571" "BRCA1 gene" (4 . "gene") (("umls_type_label" . "['Gene or Genome']") ("xrefs" . "['NCI_NCI-HGNC:HGNC:1100', 'CHV:0000031821', 'PDQ:CDR0000043111', 'MESH:D019398', 'CSP:4005-0006', 'MTH:NOCODE', 'LNC:LP36227-4', 'NCI:C17965', 'LNC:LP19666-4', 'OMIM:113705', 'HGNC:HGNC:1100']") ("id" . "UMLS:C0376571") ("umls_type" . "['T028']") ("labels" . "['gene']")))
    (robokop 26 "HGNC:1100" "BRCA1" (0 . "(\"named_thing\" \"gene\")") (("locus_group" . "protein-coding gene") ("chromosome" . "17") ("taxon" . "9606") ("gene_family" . "(\"Ring finger proteins\" \"FA complementation groups\" \"Protein phosphatase 1 regulatory subunits\" \"BRCA1 A complex\" \"BRCA1 B complex\" \"BRCA1 C complex\")") ("location" . "17q21.31") ("id" . "HGNC:1100") ("gene_family_id" . "(58 548 694 1328 1335 1336)") ("equivalent_identifiers" . "(\"UniProtKB:C9IZW4\" \"UniProtKB:E9PC22\" \"UniProtKB:A0A2R8Y7V5\" \"UniProtKB:H0Y8D8\" \"UniProtKB:E9PH68\" \"UniProtKB:K7EPC7\" \"UniProtKB:E7EQW4\" \"UniProtKB:H0Y881\" \"UniProtKB:E7EWN5\" \"UniProtKB:H0Y850\" \"UniProtKB:C6YB45\" \"UniProtKB:E7EUM2\" \"UniProtKB:A0A024R1V0\" \"HGNC:1100\" \"UniProtKB:A0A0U1RRA9\" \"UniProtKB:E7ENB7\" \"UniProtKB:K7EJW3\" \"UniProtKB:H0Y8B8\" \"UniProtKB:A0A2R8Y6Y9\" \"UniProtKB:Q5YLB2\" \"UniProtKB:P38398\" \"UniProtKB:B7ZA85\" \"UniProtKB:A0A0A0MSN1\" \"ENSEMBL:ENSG00000012048\" \"UniProtKB:Q3B891\" \"UniProtKB:G1UI37\" \"NCBIGENE:672\" \"UniProtKB:A0A2R8Y587\")")))
    (orange 32553 "NCBIGene:672" "BRCA1" (6 . "(\"gene\")") (("iri" . "http://www.ncbi.nlm.nih.gov/gene/672") ("synonym" . "(\"BRCA1/BRCA2-containing complex, subunit 1\" \"Fanconi anemia, complementation group S\" \"protein phosphatase 1, regulatory subunit 53\" \"BRCC1\" \"FANCS\" \"PPP1R53\" \"RNF53\" \"BREAST CANCER 1 GENE; BRCA1\" \"BRCA1\")") ("in_taxon" . "NCBITaxon:9606") ("same_as" . "(\"ENSEMBL:ENSG00000012048\" \"HGNC:1100\" \"OMIM:113705\" \"Orphanet:119068\")") ("provided_by" . "(\"orphanet.ttl\" \"omim.ttl\")") ("description" . "BRCA1, DNA repair associated") ("id" . "NCBIGene:672")))))


;;; Want to restrict to gene/protein/whatever
;;;
;;; This is pretty restrictive, and doesn't include 'rtx' (which seems
;;; to use 'protein' instead of gene).
(define (restrict-concept-to-gene dbname c)
  (fresh (cid cui name catid cat props)
    (== `(,cid ,cui ,name (,catid . ,cat) ,props) c)
    (membero `(,dbname (,catid . ,cat))
             `((semmed (4 . "gene"))
               (robokop (0 . "(\"named_thing\" \"gene\")"))    
               (orange (6 . "(\"gene\")"))))))

;;; pretty restrictive!
;;; what about orange?
(define (restrict-concept-to-drug dbname c)
  (fresh (cid cui name catid cat props)
    (== `(,cid ,cui ,name (,catid . ,cat) ,props) c)
    (membero `(,dbname (,catid . ,cat))
             `((semmed (5 . "chemical_substance"))
               (rtx (0 . "chemical_substance"))
               (robokop (1 . "(\"named_thing\" \"chemical_substance\")"))))))

;;; don't allow gene classes, such as "Genes", "Tumor Suppressor Genes", etc.
(define (filter-out-bogus-genes dbname c)
  (fresh (cid cui name catid cat props)
    (== `(,cid ,cui ,name (,catid . ,cat) ,props) c)
    (conde
      ((== 'semmed dbname)
       (=/= "Genes" name)
       (=/= "Homologous Gene" name)
       (=/= "Tumor Suppressor Genes" name))
      ((=/= 'semmed dbname)))))

#|
;; seems like this might be a better/more general way to determine if
;; a concept is really a class of concepts; alas, semmed's 'subclass_of'
;; seems horrible:  A subclass_of B subclass_of A
;;; So, not sure how to use this!
(define (concept-has-subclass dbname c)
  (not (null? (run 1 (o)
                (fresh (dbname eid pid eprops e)
                  (== 'semmed dbname) ;; how to handle others?
                  (== `(,dbname ,eid ,c ,o (4 . "subclass_of") ,eprops) e)
                  (edgeo e))))))
|#

#|
;;; uh oh!!  sigh
(time
 (concept-has-subclass
  'semmed
  '(74686
    "UMLS:C0376571"
    "BRCA1 gene"
    (4 . "gene")
    (("umls_type_label" . "['Gene or Genome']")
     ("xrefs" . "['NCI_NCI-HGNC:HGNC:1100', 'CHV:0000031821', 'PDQ:CDR0000043111', 'MESH:D019398', 'CSP:4005-0006', 'MTH:NOCODE', 'LNC:LP36227-4', 'NCI:C17965', 'LNC:LP19666-4', 'OMIM:113705', 'HGNC:HGNC:1100']")
     ("id" . "UMLS:C0376571")
     ("umls_type" . "['T028']")
     ("labels" . "['gene']")))))
cpu time: 1 real time: 3 gc time: 0
#t

(time
  (concept-has-subclass
    'semmed
    '(17324
      "UMLS:C0017337"
      "Genes"
      (4 . "gene")
      (("umls_type_label" . "['Gene or Genome']")
       ("xrefs"
        .
        "['CSP:1256-5501', 'NCI:C16612', 'LNC:LP32747-5', 'NCI_NCI-GLOSS:CDR0000045693', 'LNC:LP199195-1', 'NCI_CDISC:C16612', 'PDQ:CDR0000042941', 'SNMI:F-E0000', 'MESH:D005796', 'LCH_NW:sh91000344', 'NCI:TCGA', 'AOD:0000002944', 'SNOMEDCT_US:67271001', 'UWDA:74402', 'CHV:0000005419', 'HL7V3.0:GENE', 'FMA:74402', 'PSY:20820']")
       ("id" . "UMLS:C0017337")
       ("umls_type" . "['T028']")
       ("labels" . "['gene']")))))
;; cpu time: 3 real time: 3 gc time: 0
;; #t

(time
 (concept-has-subclass
  'semmed
  '(141262
    "UMLS:C1425023"
    "NGLY1 gene"
    (4 . "gene")
    (("umls_type_label" . "['Gene or Genome']")
     ("xrefs" . "['HGNC:HGNC:17646', 'OMIM:610661', 'MTH:NOCODE']")
     ("id" . "UMLS:C1425023")
     ("umls_type" . "['T028']")
     ("labels" . "['gene']")))))
;; cpu time: 1 real time: 2 gc time: 0
;; #f

(time
 (concept-has-subclass
  'semmed
  '(33093
    "UMLS:C0079427"
    "Tumor Suppressor Genes"
    (4 . "gene")
    (("umls_type_label" . "['Gene or Genome']")
     ("xrefs" . "['SNOMEDCT_US:77864004', 'NCI_NCI-GLOSS:CDR0000046657', 'MESH:D016147', 'CHV:0000015171', 'MTH:NOCODE', 'SNMI:F-E00A0', 'NCI_NCI-GLOSS:CDR0000583866', 'SNOMEDCT_US:405842004', 'LCH_NW:sh90002154', 'OMIM:MTHU005882', 'NCI:C17362', 'CSP:1256-6400']")
     ("id" . "UMLS:C0079427")
     ("umls_type" . "['T028']")
     ("labels" . "['gene']")))))
;; cpu time: 1 real time: 1 gc time: 0
;; #t
|#

;;; Filter out "Genes", "Tumor Suppressor Genes", and other losers etc.
;;; Or provide interface for hoomon to do so.
;;;
;;; Also, can end up with duplicate concepts/CUIs, due to different KGs and also different predicates;
;;; be smart about this!!
(define G1-G2-edges
  (remove-duplicates
   (run* (e)
     (fresh (dbname eid G1 G2 pid pred eprops)
       (== `(,dbname ,eid ,G1 ,G2 (,pid . ,pred) ,eprops) e)
       ;; don't want G1 and G2 to be the same gene!
       (=/= G1 G2)
       (membero `(,dbname . ,G2) G2-concepts)
       (edgeo e)
       (membero pred G-G-increases-predicate-names)
       (restrict-concept-to-gene dbname G1)
       ;; should be configurable which concepts are "bogus"
       (filter-out-bogus-genes dbname G1)))))

#|
(remove-duplicates
 (map
  (lambda (e)
    (match e
      [`(,dbname ,eid
                 (,scid ,scui ,sname (,scatid . ,scat) ,sprops)
                 (,ocid ,ocui ,oname (,ocatid . ,ocat) ,oprops)
                 (,pid . ,pred) ,eprops)
       (list scui sname pred ocui oname (pubmed-URLs-from-edge e) dbname)]))
  G1-G2-edges))
|#

;;; Can end up with duplicate concepts/CUIs, due to different KGs and also different predicates;
;;; be smart about this!!
(define G1-G2-edges-without-duplicate-dbname/scui
  (let loop ([edges G1-G2-edges])
    (match edges
      ['() '()]
      [`(,e . ,rest)
       (match e
         [`(,dbname ,eid
                    (,scid ,scui ,sname (,scatid . ,scat) ,sprops)
                    (,ocid ,ocui ,oname (,ocatid . ,ocat) ,oprops)
                    (,pid . ,pred) ,eprops)
          (let ((e^ (memf
                     (lambda (e^)
                       (match e^
                         [`(,dbname^ ,eid^
                                     (,scid^ ,scui^ ,sname^ (,scatid^ . ,scat^) ,sprops^)
                                     (,ocid^ ,ocui^ ,oname^ (,ocatid^ . ,ocat^) ,oprops^)
                                     (,pid^ . ,pred^) ,eprops^)
                          (equal? (list dbname scui) (list dbname^ scui^))]))
                     rest)))
            (if e^
                (loop rest)
                (cons e (loop rest))))])])))

(define D-G1-G2-paths
  (remove-duplicates
   (run 150 (q)
     (fresh (dbname1 eid1 D G1 pid1 pred1 eprops1 e1
                     dbname2 eid2 G2^ G2 pid2 pred2 eprops2 e2
                     cid0 cui0 name0 catid0 cat0 props0
                     cid1 cui1 name1 catid1 cat1 props1
                     cid2 cui2 name2 catid2 cat2 props2
                     cid3 cui3 name3 catid3 cat3 props3)
       (== `(,dbname1 ,eid1 ,D ,G1 (,pid1 . ,pred1) ,eprops1) e1)
       (== `(,dbname2 ,eid2 ,G2^ ,G2 (,pid2 . ,pred2) ,eprops2) e2)

       (== `(,e1 ,e2) q)

       (== `(,cid0 ,cui0 ,name0 (,catid0 . ,cat0) ,props0) D)
       (== `(,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1) G1)
       (== `(,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2) G2^)
       (== `(,cid3 ,cui3 ,name3 (,catid3 . ,cat3) ,props3) G2)
       
       (== cui1 cui2)
       (== dbname1 dbname2)
       
       (membero e2 G1-G2-edges-without-duplicate-dbname/scui)
       (edgeo e1)
       (restrict-concept-to-drug dbname1 D)
       (membero pred1 D-G-increases-predicate-names)
       ))))

(map
 (lambda (p)
   (match p
     [`(,e1 ,e2)
      (match `(,e1 ,e2)
        [`((,dbname1 ,eid1
                     (,cid0 ,cui0 ,name0 (,catid0 . ,cat0) ,props0)
                     (,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1)
                     (,pid1 . ,pred1)
                     ,eprops1)
           (,dbname2 ,eid2
                     (,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2)
                     (,cid3 ,cui3 ,name3 (,catid3 . ,cat3) ,props3)
                     (,pid2 . ,pred2)
                     ,eprops2))
         `(,name0 ,pred1 ,(pubmed-URLs-from-edge e1) ,name1 ,dbname1
                  ,pred2 ,(pubmed-URLs-from-edge e2) ,name3 ,dbname2)])]))
 D-G1-G2-paths)

#|
(time
 (remove-duplicates
   (run 1 (q)
     (fresh (dbname1 eid1 s1 o1 pid1 pred1 eprops1 e1
             dbname2 eid2 s2 o2 pid2 pred2 eprops2 e2
             cid0 cui0 name0 catid0 cat0 props0
             cid1 cui1 name1 catid1 cat1 props1
             cid2 cui2 name2 catid2 cat2 props2
             cid3 cui3 name3 catid3 cat3 props3)
       (== `(,dbname1 ,eid1 ,s1 ,o1 (,pid1 . ,pred1) ,eprops1) e1)
       (== `(,dbname2 ,eid2 ,s2 ,o2 (,pid2 . ,pred2) ,eprops2) e2)

       (== `(,name0 ,pred1 ,name1 ,pred2 ,name3 ,dbname1) q)

       (== `(,cid0 ,cui0 ,name0 (,catid0 . ,cat0) ,props0) s1)
       (== `(,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1) o1)
       (== `(,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2) s2)
       (== `(,cid3 ,cui3 ,name3 (,catid3 . ,cat3) ,props3) o2)
       
       (== cui1 cui2)
       (== dbname1 dbname2)
       
       (membero e2 G1-G2-edges-without-duplicate-dbname/scui)
       (edgeo e1)
       (restrict-concept-to-drug dbname1 s1)
       (membero pred1 D-G-increases-predicate-names)
       ))))
|#

#|
(time
 (remove-duplicates
   (run 1 (q)
     (fresh (dbname1 eid1 s1 o1 pid1 pred1 eprops1 e1
             dbname2 eid2 s2 o2 pid2 pred2 eprops2 e2
             cid0 cui0 name0 catid0 cat0 props0
             cid1 cui1 name1 catid1 cat1 props1
             cid2 cui2 name2 catid2 cat2 props2
             cid3 cui3 name3 catid3 cat3 props3)
       (== `(,dbname1 ,eid1 ,s1 ,o1 (,pid1 . ,pred1) ,eprops1) e1)
       (== `(,dbname2 ,eid2 ,s2 ,o2 (,pid2 . ,pred2) ,eprops2) e2)

       (== `(,name0 ,pred1 ,name1 ,pred2 ,name3 ,dbname1) q)

       (== `(,cid0 ,cui0 ,name0 (,catid0 . ,cat0) ,props0) s1)
       (== `(,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1) o1)
       (== `(,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2) s2)
       (== `(,cid3 ,cui3 ,name3 (,catid3 . ,cat3) ,props3) o2)
       
       (== cui1 cui2)
       (== dbname1 dbname2)
       
       (membero e2 G1-G2-edges)
       (edgeo e1)
       (restrict-concept-to-drug dbname1 s1)
       (membero pred1 D-G-increases-predicate-names)
       ))))

(time
 (remove-duplicates
   (run 1 (q)
     (fresh (dbname1 eid1 s1 o1 pid1 pred1 eprops1 e1
             dbname2 eid2 s2 o2 pid2 pred2 eprops2 e2
             cid0 cui0 name0 catid0 cat0 props0
             cid1 cui1 name1 catid1 cat1 props1
             cid2 cui2 name2 catid2 cat2 props2
             cid3 cui3 name3 catid3 cat3 props3)
       (== `(,dbname1 ,eid1 ,s1 ,o1 (,pid1 . ,pred1) ,eprops1) e1)
       (== `(,dbname2 ,eid2 ,s2 ,o2 (,pid2 . ,pred2) ,eprops2) e2)

       (== `(,name0 ,pred1 ,name1 ,pred2 ,name3 ,dbname1) q)

       (== `(,cid0 ,cui0 ,name0 (,catid0 . ,cat0) ,props0) s1)
       (== `(,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1) o1)
       (== `(,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2) s2)
       (== `(,cid3 ,cui3 ,name3 (,catid3 . ,cat3) ,props3) o2)
       
       (== cui1 cui2)
       (== dbname1 dbname2)
       
       (membero e2 G1-G2-edges)
       (edgeo e1)
       (restrict-concept-to-drug dbname1 s1)
       (membero pred1 D-G-increases-predicate-names)
       ))))
|#

#|
;;; slow query 1
;;; cpu time: 24794 real time: 25191 gc time: 3238
(time
 (length
  (remove-duplicates
   (run 1 (path)
     (fresh (dbname1 eid1 s1 o1 pid1 pred1 eprops1 e1
                     dbname2 eid2 s2 o2 pid2 pred2 eprops2 e2
                     cid1 cui1 name1 catid1 cat1 props1
                     cid2 cui2 name2 catid2 cat2 props2)
       (== `(,dbname1 ,eid1 ,s1 ,o1 (,pid1 . ,pred1) ,eprops1) e1)
       (== `(,dbname2 ,eid2 ,s2 ,o2 (,pid2 . ,pred2) ,eprops2) e2)
       (== `(,e1 ,e2) path)

       (== `(,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1) o1)
       (== `(,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2) s2)
       (== cui1 cui2)
       (membero e2 G1-G2-edges)
       (edgeo e1)
       (restrict-concept-to-drug dbname1 s1)
       (membero pred1 D-G-increases-predicate-names)
       )))))

;;; swap two goals in the query above to make it suuuupeeer slow
(time
  (length
    (remove-duplicates
      (run 1 (path)
        (fresh (dbname1 eid1 s1 o1 pid1 pred1 eprops1 e1
                        dbname2 eid2 s2 o2 pid2 pred2 eprops2 e2
                        cid1 cui1 name1 catid1 cat1 props1
                        cid2 cui2 name2 catid2 cat2 props2)
          (== `(,dbname1 ,eid1 ,s1 ,o1 (,pid1 . ,pred1) ,eprops1) e1)
          (== `(,dbname2 ,eid2 ,s2 ,o2 (,pid2 . ,pred2) ,eprops2) e2)
          (== `(,e1 ,e2) path)

          (== `(,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1) o1)
          (== `(,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2) s2)
          (== cui1 cui2)
          (membero e2 G1-G2-edges)
          (restrict-concept-to-drug dbname1 s1)
          (edgeo e1)
          (membero pred1 D-G-increases-predicate-names)
          )))))
|#






#|
(time
  (length
    (remove-duplicates
      (run 1 (path)
        (fresh (dbname1 eid1 s1 o1 pid1 pred1 eprops1 e1
                        dbname2 eid2 s2 o2 pid2 pred2 eprops2 e2
                        cid1 cui1 name1 catid1 cat1 props1
                        cid2 cui2 name2 catid2 cat2 props2)
          (== `(,dbname1 ,eid1 ,s1 ,o1 (,pid1 . ,pred1) ,eprops1) e1)
          (== `(,dbname2 ,eid2 ,s2 ,o2 (,pid2 . ,pred2) ,eprops2) e2)
          (== `(,e1 ,e2) path)

          (== `(,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1) o1)
          (== `(,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2) s2)
          (== cui1 cui2)
          (membero e2 G1-G2-edges)
          (restrict-concept-to-drug dbname1 s1)
          (edgeo e1)
          (membero pred1 D-G-increases-predicate-names)
          )))))
|#

#|
(define D-G1-G2-paths
  (remove-duplicates
   (run* (path)
     (fresh (dbname1 eid1 s1 o1 pid1 pred1 eprops1 e1
             dbname2 eid2 s2 o2 pid2 pred2 eprops2 e2
             cid1 cui1 name1 catid1 cat1 props1
             cid2 cui2 name2 catid2 cat2 props2)
       (== `(,dbname1 ,eid1 ,s1 ,o1 (,pid1 . ,pred1) ,eprops1) e1)
       (== `(,dbname2 ,eid2 ,s2 ,o2 (,pid2 . ,pred2) ,eprops2) e2)
       (== `(,e1 ,e2) path)

       (== `(,cid1 ,cui1 ,name1 (,catid1 . ,cat1) ,props1) o1)
       (== `(,cid2 ,cui2 ,name2 (,catid2 . ,cat2) ,props2) s2)
       (== cui1 cui2) ;; tie the KG's together
       
       (membero e2 G1-G2-edges)
       (edgeo e1)
       (membero pred1 D-G-increases-predicate-names)
       (restrict-concept-to-drug dbname1 s1)))))
|#



#|
Exclude list?

(160755
 "UMLS:C1334043"
 "Homologous Gene"
 (4 . "gene")
 (("umls_type_label" . "['Gene or Genome']")
  ("xrefs" . "['NCI:C28709', 'MTH:NOCODE']")
  ("id" . "UMLS:C1334043")
  ("umls_type" . "['T028']")
  ("labels" . "['gene']")))

|#

#|
(~name*-concepto ~name* concept)

(~cui-concepto ~cui concept)

(~categoryo ~category-name category)

(~predicateo ~predicate-name predicate)

(sort-paths paths)

(pmid-edgeo pmid edge)

(pubmed-URLs-from-edge edge)

(define predicates
  (sort (remove-duplicates
         (time (run* (predicate)
                 (fresh (dbname pid c)
                   (membero c selected-concepts)
                   (concept-predicateo c `(,dbname ,pid . ,predicate))))))
        string<?))

(run* (q) (fuzzy-concepto current-name q))

(remove-duplicates
 (run 50 (s-with-dbname) ;; 50 should probably be a parameter ;
   (fresh (o-with-dbname dbname o s eid pid eprops e)
     (membero o-with-dbname ans)
     (== `(,dbname . ,o) o-with-dbname)
     (== `(,dbname ,eid ,s ,o (,pid . "subclass_of") ,eprops) e)
     (== `(,dbname . ,s) s-with-dbname)
     (edgeo e))))

(set! paths
      (remove-duplicates
       (append paths
               (run* (q)
                 (fresh (e dbname eid x o pid pred eprops)
                   (== (list `(,dbname ,eid ,x ,o (,pid . ,pred) ,eprops)) q)
                   (== `(,dbname . ,x) selected-X)
                   (== `(,dbname ,eid ,x ,o (,pid . ,pred) ,eprops) e)
                   (membero `(,dbname . ,o) concept-2*)
                   (membero pred atomic-predicate-2*)
                   (edgeo e))))))

(set! paths
      (remove-duplicates
       (append paths
               (run* (q)
                 (fresh (e dbname eid s x pid pred eprops)
                   (== (list `(,dbname ,eid ,s ,x (,pid . ,pred) ,eprops)) q)
                   (== `(,dbname . ,x) selected-X)
                   (== `(,dbname ,eid ,s ,x (,pid . ,pred) ,eprops) e)
                   (membero `(,dbname . ,s) concept-1*)
                   (membero pred atomic-predicate-1*)
                   (edgeo e))))))

(run* (q)
  (fresh (e dbname eid x o pid pred eprops)
    (== (list `(,dbname ,eid ,x ,o (,pid . ,pred) ,eprops)) q)
    (== `(,dbname . ,x) selected-X)
    (== `(,dbname ,eid ,x ,o (,pid . ,pred) ,eprops) e)
    (membero `(,dbname . ,o) concept-2*)
    (membero pred atomic-predicate-2*)
    (edgeo e)))

(set! all-X-concepts-with-edges
      (remove-duplicates
       (append all-X-concepts-with-edges
               (run* (q)
                 (fresh (dbname eid s o pid pred eprops e)
                        ;; TODO FIXME -- epropos may contain pubmed ids--how to extract it, or other evidence? ;
                   (== (list dbname s (list eprops) (list e)) q)
                   (== `(,dbname ,eid ,s ,o (,pid . ,pred) ,eprops) e)
                   (membero `(,dbname . ,o) concept-2*)
                   (membero pred atomic-predicate-2*)
                   (edgeo e))))))

(filter ;; only include concepts with at least one predicate ;
 (lambda (x)
   (match x
     [`(,dbname . ,concept)
      (let ((preds
             (run 1 (pred)
               (fresh (o s eid eprops e)
                 (case edge-type
                   [(out-edge) (== concept s)]
                   [(in-edge) (== concept o)])
                 (== `(,dbname ,eid ,s ,o ,pred ,eprops) e)
                 (edgeo e)))))
        (not (null? preds)))]))
 ans)

|#