Ali-U-Net: A Convolutional Transformer Neural Net for Multiple Sequence Alignment of DNA Sequences. A proof of concept
Arsic, P.; Mayer, C.
Show abstract
We report a convolutional transformer neural network that is capable of aligning multiple nucleotide sequences. The neural network is based on the U-Net commonly used in image segmentation which we employ to transform unaligned sequences to aligned sequences. For alignment scenarios our Ali-U-Net neural network has been trained on, it is in most cases more accurate than programs such as MAFFT, T-Coffee, MUSCLE, and Clustal Omega, while being considerably faster than similarly accurate programs on a single CPU core. Limitations are that the neural network is still trained specifically for certain alignment problems and can perform poorly for gap distributions it has not seen before. Furthermore, the algorithm currently works with fixed-size alignment windows of 48x48 or 96x96 nucleotides. At this stage, we view our study as a proof of concept, confident that the present findings can be extended to larger alignments and more complex alignment scenarios in the near future.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- TrieDedup: A fast trie-based deduplication algorithm to handle ambiguous bases in high-throughput sequencing 95%
- Read-SpaM: assembly-free and alignment-free comparison of bacterial genomes with low sequencing coverage 94%
- SPAligner: alignment of long diverged molecular sequences to assembly graphs 94%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.