qdata / fastsk Goto Github PK

Hi,

I am trying to use FastSK, but a little confused about how to prepare the training dataset.

For example, I want to train a model of ATAC-seq in K562 cells, should the positive data be real ATAC-seq data and negative data be randomized seqs?

Does 0 represent positive seqs and 1 represent randomized negative seqs ?

Thanks a lot.

Is there a way to retrieve a list of the most contributing features (i.e., k-mer sequences) and their associated weights? I have attempted to pull them by using the sklearn.feature_extraction module to no avail. However, I did notice (with my incredibly limited familiarity with c++) that there appears to be the tokenized feature array objects being passed to fastsk_kernel.cpp. Is this a potential source of extraction? Any help would be greatly appreciated!

Hi, if we have continuous data rather than a binary classification task, is there any easy way to run this as a support vector regression (SVR) problem rather than typical SVM? Or is this package intended purely as a classifier? Thanks!

I ran iGaKCo as below and it core dumped:

$ ./iGakco -g 10 -m 8 -k kernel.txt -h 1 temp.fasta temp.fasta dict.txt labels.txt 
Input file : temp.fasta
Dictionary size = 5 (+1 for uknown character)
Reading temp.fasta
Read 2 strings of length = 12
numF=6, sumLen=24
(10,2): 6 features
Weights (hm):45 36 28 21 15 10 6 3 1 
Computing mismatch profiles using 8 threads...
*** Error in `./iGakco': free(): invalid size: 0x00007f40e4000c70 ***
*** Error in `./iGakco': free(): invalid size: 0x00007f40ec000c70 ***
*** Error in `./iGakco': free(): invalid size: 0x00007f40e0000c70 ***
*** Error in `./iGakco': free(): invalid size: 0x00007f40d4000c70 ***
*** Error in `./iGakco': free(): invalid size: 0x00007f40d8000c70 ***
======= Backtrace: =========
======= Backtrace: =========
======= Backtrace: =========
======= Backtrace: =========
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x8037a)[0x7f40f373737a]
/lib/x86_64-linux-gnu/libc.so.6(+0x777e5)[0x7f40f372e7e5]
*** Error in `./iGakco': free(): invalid size: 0x00007f40c8000c70 ***
Aborted (core dumped)

Input file looks like this:

>1
ACGACGTGACGA
>0
AGTCCCAAATCC

dict.txt is:

A
C
G
T