Previous studies have shown that information about sequence conservation can improve transcription factor binding site predictions [3], [11], but we wanted to better understand how conservation improves accuracy.
We tested our benchmark on five methods, three of which use sequence conservation to improve prediction accuracy.
The weighted sum (WS) method is the most straight forward of the methods we tested that use sequence conservation to improve accuracy.
SLiMFinder uses global or local sequence conservation to improve confidence in motif predictions [9], [24]; DILIMOT takes into account conservation of the motif in orthologs as part of the scoring scheme [10].
Previous studies have shown that sequence conservation can improve transcription factor binding site predictions [ 11, 12].
RNA editing events are observed preferentially in protein-coding regions of land plant mitochondria at first and second positions in codons and tend to improve the level of protein sequence conservation.
Classification of TNF and USP, families with low sequence conservation, is also much improved in the rewired network, showing that rewiring restores missing edges due to remote homology, as well as removing edges due to domain chaining.
More complete genome assemblies, or improved and more sensitive algorithms, might detect sequence conservation at these enhancers; while we cannot rule out this possibility, our results were obtained with the best assemblies and computational tools available at the time.
ExAlt identifies the structure of most alternatively spliced exons in the test set and cross-species sequence conservation is shown to improve the precision of predictions.
Second, we test this benchmark on a small set of methods to investigate the effects of using an alternative to the common position-weight matrix (PWM) motif representation, and of using sequence conservation across related genomes to improve accuracy.
