Given, say, 150 different fasta sequences, we need to find an optimal way to bin them where each bin is around 2kb.

Thus, we find ourselves with a variant of the Knapsack problem. The code for this tool was heavily borrowed from https://github.com/handcraftsman/GeneticGo's "samples/ukp/rosetta.go" file.

Normally in UKP, you're trying to find the maximal value of a single bin, given some contraints like weight/volume and some motivators like "value".

We abuse the heck out of this (in order to have to write less new code) by writing a fitness function which:

• treats "count" (number of objects per bin in UKP) as the bin number
• biasing heavily towards low bin numbers (to force multiple items into bins together)
• biasing against bins which are far from our target size (similar to weight/volume)
• biasing heavily against solutions which do not use all of the available resources (i.e. sequences)

$ go run bin_fasta.go -lengthTable test.tsv -targetLength=2000 > out.tsv

This will take the features defined in test.tsv and their lengths in column 2, and try to group them such that each grouping is approximately 2kb. See -help for options.