This is a README file, there are many like it, but this is my own.

Start with an explanation of the database and its schemas.
The primary key for everything lies in the genome table.

genome:
Id Acccession Species Genename Version Comment Mrna_seq Protein_seq Orf_start Orf_stop Lastupdate
  Id is auto-incremented.  Genename must be parsed out of the input database and so is not as
  reliable as comment.  mrna_seq is the full sequence given.  protein_seq comes from the cds
  annotation.  orf_start and orf_stop define what range we will look at later.
Problems:
  1)  This table is filled with 1 full copy of the genomic sequence (mrna_seq) for every
      orf in the sequence, this is annoying for viral genomes.

gene_info:
  Same as GENOME except text information about each gene, indexed for fast searching.

queue:
Id Genome_id Public Params Out Done
  Each id is unique, but the genome_id will be used to pull the sequence
  SELECT id, genome_id, params where public='1' order by rand() limit 1
  SELECT id, genome_id, params where done='0' order by rand() limit 1
  should pull the parameters and genome for a random one
  UPDATE queue set out='1' where id='xxx'
  UPDATE queue set done='1' where id='xxx'
  should tell the database that a given sequence is finished

mfe_$species:
Id Genome_id Species Accession Start Slipsite Seqlength Sequence Output Parsed Mfe Pairs Knotp Barcode Lastupdate
  Same ideas apply.  The only time the Accession should be duplicated is when you have
   a)  Multiple slippery sites per transcript
   b)  Multiple ORFs per transcript

boot_$species:
ID Genome_id Species Accession Start Iterations Rand_method Mfe_method Mfe_mean Mfe_Sd Mfe_Se Pairs_mean Pairs_sd Pairs_se Mfe_values Lastupdate

landscape_$species:
The largest tables in the database, contain the mfe data for entire sequences.

overlap:
Looks for + and - 1 frameshift products

agree:
Generates statistics about how well nupack,pknots, and hotknots agree for each sequence

index_stats:
Generates statistics used for the web site.