Deep network for retroelement insertion motifs

[avifriedlich]

Retroelement insertion has been a major contributor to genomic evolutuion across mammalian species, and Identifying sequence patterns targeted for retroelement insertion is a tractable problem. I've visualized these non-linear patterns by dot-plot and made some pictures (eg http://www.cell.com/pictureshow/bestof2013), but machine learning is what will solve the problem and open major new roads in genomics and it requires a team effort.

For example, the 300 n.t. AluY is unique to the human genome where it is inserted 8 million times in intronic A/T rich areas.

If anyone with expertise in convolutional design feels the same way about it, I would be delighted to curate data in silico and assist in deep network development.

Avi Friedlich MD