search

tderrien / FEELnc

FEELnc : FlExible Extraction of LncRNA
GNU General Public License v3.0
79 stars 28 forks source link

codpot performance #33

Closed Simon602 closed 5 years ago

Simon602 commented 5 years ago

Dear All

in the codpot.pl, I used the lncRNAs sequence to train the algorithm but the in the graph ROC curves there is a shifting in mRNA specificity curves when I launched it with -l flag to train algorithm with lncRNAs fasta sequence compare to --mode=shuffle or intergenic flags, why coding probability for mRNA specificity curve started at 0.2 ? according to the curves which method is better, training with lncRNA.fasta or --mode=shuffle ? Thanks candidate_lncrna_traning_mode groc codpot_shuffle_flag

tderrien commented 5 years ago

HI @Simon602 ,

Regarding your first question, the mRNA specificity actually starts at ~0 but the y-axis limits is zoomed to 0,5->1.

according to the curves which method is better, training with lncRNA.fasta or --mode=shuffle ?

It really depends on the confidence you have in your lncRNA training file. If your are not sure about their robustness, you can indeed use the shuffle method (which performs better than intergenic in our analysis) Bw,

Thomas