Neural network input representations that produce accurate consensus sequences from DNA fragment assemblies

Motivation: Given inputs extracted from an aligned column of DNA bases and the underlying Pet-kin Elmer Applied Biosystems (ABI) fluorescent tr-aces, our goal is to train a neural network to determine correctly the consensus base for the column. Choosing an appropriate network input representation i s critical to success in this task. We empirically compare five representat ions; one uses only base calls and the others include trace information. Results: We attained the most accurate results from networks that incorpora te trace information into their input representations. Based on estimates d erived from using 10-fold cross-validation, the best network topology produ ces consensus accuracies ranging from 99.26% to >99.98% for coverages from two to six aligned sequences. With a coverage of six, it makes only three e rrors in 20 000 consensus calls. In contrast, the network that only uses ba se calls in its input representation has over double that error rate: eight errors in 20 000 consensus calls.