samstudio8 / goldilocks Goto Github PK

Redundancy in the genetic code means that it is possible to represent 64 different amino acids even though there are only 21 used. This has resulted in many codons being used for each amino acid. Some species prefer to use one codon over another which results in a species specific codon usage profile.

Can Goldilocks be used to analyse the frequency that a particular codon is used in a genome?

nope.fa

>NO
>OP

nope.fa.idx

NO      0       4       -1      -1
OP      0       8       -1      -1

nope.py

from goldilocks.goldilocks import Goldilocks
from goldilocks.strategies import NucleotideCounterStrategy

data = {
    "seq": {
        "idx": "nope.fa.idx"
    }
}

g = Goldilocks(NucleotideCounterStrategy(["A"]), data, length=3, stride=1, is_faidx=True) 
g.query("mean", track="A", limit=5).export_meta(sep="\t")

File "/home/sam/Projects/Packages/goldilocks/goldilocks/goldilocks.py", line 525, in __apply_filter_func
    mean_percentile = (len(track_scores[track_scores <= 0]) / float(len(track_scores))) * 100
ZeroDivisionError: float division by zero

• Remove files that are not required for distribution
• Ensure all options are set as needed
• Remove reliance on terrible Makefile
• Better options for compilation of documentation

A comment raised on a reddit post about Goldilocks points out that some of the operations performed by the tool might well serve as standalone tools to be used in UNIX fashion.

At the very least, supporting streamed input could be a nice idea.

I'd like to be able to supply individual files rather than one large FASTA file. For example:

sequence_data = { "mysample": {
        "chr1": {"file": "./chr1.fna.fai"},
        "chr2": {"file": "./chr2.fna.fai"},
        "chr3": {"file": "./chr3.fna.fai"},
        "chr4": {"file": "./chr4.fna.fai"},
        "chr5": {"file": "./chr5.fna.fai"},
}}
g = Goldilocks(GCRatioStrategy(), sequence_data, length="500K", stride="100K", is_faidx=True)
g.plot(group="mysample", title="GC Content")

Goldilocks should output results in a more useful format than tabular, BED seems a good bet.

The PositionCounterStrategy uses the default track counter, which is then used later for calculating summary statistics. This has the side effect of reporting G times more variants than there actually are.

It could be worth investigating providing a mechanism for post-census functions to be called on the count matrix once a census has been completed. This could allow for a GCRatioStrategy to just be a wrapper of the NucleotideCounterStrategy that has a post-census callback to calculate and return the necessary ratios.