@article{fdi:010090309,
  title = {{G}enomic object detection : an improved approach for transposable elements detection and classification using convolutional neural networks},
  author = {{O}rozco-{A}rias, {S}. and {L}opez-{M}urillo, {L}. {H}. and {P}iña, {J}. {S}. and {V}alencia-{C}astrillon, {E}. and {T}abares-{S}oto, {R}. and {C}astillo-{O}ssa, {L}. and {I}saza, {G}. and {G}uyot, {R}omain},
  editor = {},
  language = {{ENG}},
  abstract = {{A}nalysis of eukaryotic genomes requires the detection and classification of transposable elements ({TE}s), a crucial but complex and time-consuming task. {T}o improve the performance of tools that accomplish these tasks, {M}achine {L}earning approaches ({ML}) that leverage computer resources, such as {GPU}s ({G}raphical {P}rocessing {U}nit) and multiple {CPU} ({C}entral {P}rocessing {U}nit) cores, have been adopted. {H}owever, until now, the use of {ML} techniques has mostly been limited to classification of {TE}s. {H}erein, a detection-classification strategy (named {YORO}) based on convolutional neural networks is adapted from computer vision ({YOLO}) to genomics. {T}his approach enables the detection of genomic objects through the prediction of the position, length, and classification in large {DNA} sequences such as fully sequenced genomes. {A}s a proof of concept, the internal protein-coding domains of {LTR}-retrotransposons are used to train the proposed neural network. {P}recision, recall, accuracy, {F}1-score, execution times and time ratios, as well as several graphical representations were used as metrics to measure performance. {T}hese promising results open the door for a new generation of {D}eep {L}earning tools for genomics. {YORO} architecture is available at https://github.com/simonorozcoarias/{YORO}.},
  keywords = {},
  booktitle = {},
  journal = {{P}los {O}ne},
  volume = {18},
  numero = {9},
  pages = {e0291925 [23 p.]},
  ISSN = {1932-6203},
  year = {2023},
  DOI = {10.1371/journal.pone.0291925},
  URL = {https://www.documentation.ird.fr/hor/fdi:010090309},
}