IREs Conclusion Page

CONCLUSIONS

During this project we had to deal with different dilemmas that would have influence on our results. An outline containing these different points is shown as follows:

As our project was based on cDNA analysis we didn't expect to find genes in reverse strand. Despite this, we checked it and realized that there were exon predictions in reverse strand with good scores. Therefore we forced the predictions in order to reduce the number of reverse genes predicted by Geneid. In spite of this, there were still reverse genes although the number was lower than before. At this point, we run a BlastP search on those reverse exons having a score higher than 20 and we figured out that most of them were real mouse proteins. This evidences indicated that not all the cDNAs were cloned in forward strand as the article said.
Bearing in mind all these evidences, we decided to make our predictions in forward because it was easier to work only in forward and the number of genes in reverse was not very high.
We used two different ways of validating our results: the first one correpond to Blast and the second one was based on a thermodynamic filter.

BLAST: The sense of running a Blast between mouse and human was supported by comparisons of animal IRE sequences revealing that the conservation of sequence identity is much higher (>90% identity) between species for the same mRNA than between different mRNAs in the same species (36%-85% identity). The chosen expected value was 0.1 due to the fact that databases were such large and the predicted IRE sequences were very short.

THERMODYNAMIC FILTER: The steps of the thermodynamic filter were the estimation of the minimum Gibbs free energy, a -2 cut off selection and checking if the pool had IRE structure. Although this is a good validation way, we think it could be better to invert the process as follows: first, we should force every sequence to adopt IRE structure and then choose those overcoming our minimum Gibbs free energy cut off. This would improve our work because we already know which kind of interactions exist in IRE structure and not always the structure with the minimum Gibbs energy is what we find in nature (as the first way of acting supposes).

After the two validations, a BlastP was run and the results were grouped in three sets containing known or similar to known proteins:

Related to iron metabolism: 11 proteins.
Without any relation to iron metabolism: 27 proteins.
Hypothetical proteins (without functional information): 7 proteins.

The last group was considered the most important because it could help us to characterize unkown proteins. These proteins seem to be real due to their structure is supported by the thermodynamic filter and they have human homologues. Despite this, more exhaustive studies (as for example, computational, "in vitro" and "in vivo" studies) would be required to confirm that these proteins contain a secondary IRE structure and that they are real proteins.