I broke everything
Don't try to run all the code now!
If you insist: python3 train.py --fasta data/Chr14.fasta --gff data/Chr14.gff --mode hasRNA python3 tester.py --fasta data/TTC5.fasta And it works much better in the case of Covid instead of human genome
Compared with ver.0.1, we need 2 major changes
1st: More generalized frame -Anze Have classes and functions only defined like: def functionOne (input): return output As the generalized frame defined, we will populate the classes and functions with codes in ver.0.1
2nd: New classifiers Someway somehow utilize gapped-kmer method with part of the seq being constrained(if necessary) 1st question: Do an estimation on required computational power if we use the method:
一：
Given key (if there is one)
vectors = []
for feature f:
for all observed motif m in look_up_dict[f]:
a = a coefficient associated with how many times m has been observed in training set
v = # of shared gapped k-mer(m, input) * a
vectors.append(v)
########### Method 1 should works! ##########
For start site at least.
Tested with estimation.py
But as l and k increase, performance could be affected
Command: python3 estimation.py --fasta data/Cov19/LC593801.1.fasta --dict chr1/Jeanne.js
#############################################
二：
Given key (if there is one)
vectors = []
for feature f:
m = most commonly observed motif in look_up_dict[f]
a = a coefficient associated with how many times m has been observed in training set
v = # of shared gapped k-mer(m, input) * a
vectors.append(v)
Tune the parameters little bit and try to find how performances could be changed and where is the limit: -Kano? If every thing works fine with L = 50 and k = 6, then we are all good! What if we combine both method through only use top 10% of the commonly seen motifs What if we have larger k values like k >= 6 What if we have each motif m being longer like len(m)>=10