VectorBase Anopheles+Aedes locus counts
  Ano. gambiae    12994 pc,  -- nc (AgamP4.3)
  Ano. funestus   13471 pc, 412 nc, 161 alts (AfunF1.3)
  Ano. albimanus  12085 pc, 424 nc, 110 alts (AalbS1.3)
  Aedes aegypti   15696 pc, 1680 nc, 1462 alts (AaegL3.3, )
                   pc= protein coding, nc= non-coding,
  
Evigene reconstruction of Anopheles+Aedes genes
  Ano. funestus   27848 loci
  Ano. albimanus  29310 loci
  Aedes aegypti   68737 loci

Q: Way too many genes there, Don, what's wrong?
A: Classify those loci with gene characteristics ..
   Conservatively classified coding loci are comparable in count to VectorBase coding set,
   additional characteristics suggest additional coding and non-coding loci.     

Classified Evigene reconstruction of Anopheles+Aedes genes
  Ano. funestus   14053..23467 pc, 2665..8303 nc loci, 29880 alts at 10653 loci
  Ano. albimanus  12000..18000 pc, nnnn..8812 nc loci, 38089 alts at 10522 loci [classify in progress] 
  Aedes aegypti   20546..23000 pc, 6344..24000 nc loci, 86437 alts at 17439 loci [classify in progress]

Don Gilbert, 2016.05.06
-----------

Evigene reconstruction of Anopheles gene loci
http://arthropods.eugenes.org/EvidentialGene/arthropods/mosquito/
  funestus  27,848 total
  
  funestus loci classified by gene characteristics
    a. insect prot homology          13201
      Of these, 40% improve homology, 48% equal, vs Vectorbase Ano.funestus loci
        
    b. related species conservation   8981  (not in a., Ano.minimus+culicifacies for Ano.fun)
      b1. coding-conservation          852  (signif Ka/Ks)
      b2. sequence-aligned            8129  (not signif Ka/Ks)
    
    c. transposon/rRNA                150  
       [excludes 440 with a.insect ho, ambiguous TE call, eg AGAP028564-PA]
    
    d. coding-potential (not a,b,c; calculated)
      d1. good coding           1285..3693  (with..without b2)
      d2. poor/non-coding       2665..8303  (w/o b2, 426 ambiguous) 

    e. contaminant/artifact   (partly in c, unfinished)

    Total classed : 26,282  
      coding.1    : 14,053 = 13201.a + 852.b1
      coding.2    : 17,746 = 13201.a + 852.b1 + 3693.d1 
      coding.3    : 23,467 = 13201.a + 8981.b1,2 + 1285.d1  
      noncoding   : 2,665..8,303 (d2)
      [remainder 1566 are what? not in ho/noho id lists]
   -----------------------     
        
  albimanus 29,310  total
  albimanus loci classified by gene characteristics
    a. insect prot homology         11649
        Of these, 41% improve homology, 38% equal, vs Vectorbase ano.albimanus loci

    b. related species conservation   nnnn  (not a.)
      b1. coding-conservation          nnn  (signif Ka/Ks)
      b2. sequence-aligned            nnnn  (not signif Ka/Ks)
      
    c. transposon             ~150

    d. coding-potential (includes a,b,c)
      d1. good coding       18731
      d2. poor/non-coding    8812,  1767 ambiguous
      
    e. contaminant/artifact   not done yet
   -----------------------     

Evigene reconstruction of Aedes gene loci
http://arthropods.eugenes.org/EvidentialGene/arthropods/mosquito/aedes_aegypti/
  aegypti 68,737 total
    good 30018, subset has insect homology and/or multiple introns on chromosomes
    poor 38725, subset lacks both homology and introns
    
  a.good, insect prot homology      20546
     Of these, 44% improve homology, 45% equal, vs Vectorbase Aedes.aegypti loci
  a.poor, insect prot homology     0
  
  b. related species conservation   (includes a.)
      b1.good, coding-conservation   20368 (sig Ka/Ks)
      b2.good, sequence-aligned       4921 (ns Ka/Ks)
      b1.poor, coding-conservation    5326 (sig Ka/Ks)
      b2.poor, sequence-aligned      13816 (ns Ka/Ks)
      
  c.good, transposons/rRNA          497..1036 (ambiguous a.insect ho vs transposons)
  c.poor, transposons/rRNA          1573 

  d. coding-potential (includes a,b,c)
    d1.good, good coding          22483
    d2.good, poor/non-coding       6344  (1235 ambiguous)
    d1.poor, good coding          12530..18012
    d2.poor, poor/non-coding      19267..24749  (1440 ambiguous)

  e. contaminant/artifact       not done yet
---------------

Coding+sequence conservation is found in related species genomes, not in related spp gene models.
A classifier for gene loci is based on measurable attributes of genes, generally
all measured for all loci.  Hierarchical classification rules then apply those measures
in order of reliability (which can include complex weightings, but simplify to):
  a. protein orthology/homology to reference + related species genes
    (i.e., protein matches well to reliably known protein)
    > remainder: b/c/e weighted character discrimination to b,c,e categories 
      (i.e., scores b/c/e = 1/0/0 > b.category, scores 0/1/0 > c, 
        scores 0.5/0.5/0.5 > ambiguous depends on reliability of scoring)
     > remainder: d coding potential character (ie. looks like coding or looks like non-coding)

Other classifying rules, and refinements, are possible (what does UniProt use?)
---------------