Entering edit mode
mauede@alice.it
▴
870
@mauedealiceit-3511
Last seen 10.2 years ago
You must think that I am pedantic and stubborn ... well, that's me
...
I put the pieces together as you painstankingly taught me to do.
You have already stressed that the following Fasta files contains
EXPERIMENTALLY VALIDATED miRNAs.
"ftp://ftp.sanger.ac.uk/pub/mirbase/sequences/CURRENT/mature.fa.gz"
"ftp://ftp.sanger.ac.uk/pub/mirbase/sequences/CURRENT/maturestar.fa.gz
"
However, if I got it right, the Homo Sapiens PREDICTED miRNAs and
genes dataset from Ensembl that looks like the following, contains
associations (miRNA,gene) that for sure have been PREDICTED but not
necessarily EXPERIMENTALLY VALIDATED. Am I right ?
GROUP SEQ METHOD FEATURE CHR START
END STRAND PHASE SCORE PVALUE_OG TRANSCRIPT_ID EXTERNAL_NAM
1 Similarity mmu-miR-707 miRanda miRNA_target 2 120824620
120824640 + . 15.3548 2.79654e-02 ENST00000295228 INHBB
2 Similarity hsa-miR-647 miRanda miRNA_target 2 120824263
120824281 + . 16.3205 3.70140e-06 ENST00000295228
INHBB
Therefore my data blocks associate miRNAs with a number of genes
3'UTR regardless of their VALIDATION. Presumably just a subset of all
predicted
are actually VALIDATED.
This is no time or energy wasted because this bulk miRNA --> genes
3'UTR data will be very useful for testing the program that is being
developed.
However, getting just the VALIDATED miRNA --> genes 3'UTR is step-1 to
train the algorithm.
A Bioinformatic guy, who is very good with Genetic and Evolutionary
algorithms, is developing a program aimed at predicting miRNA targets.
>From my understanding, these algorithms are similar to Neural
Networks, thus they have to be trained first, then tested on a
supervised set.
If these two stages are performed correctly, the algorithm should then
discriminate on the basis of rules learnt from the training.
Bottomline, I guess I can't spare myself from dealing with miRecords
XLS file as well ....
Basically I should use the miRecords VALIDATED data to screen the
information I am getting through the procedure your taught me.
The downside of this further mapping is that miRecords apparently does
not offer an ftp or any other automatic data transfer protocol.
Kind regards,
Maura
tutti i telefonini TIM!
[[alternative HTML version deleted]]