| CPC G16B 30/10 (2019.02) [G16B 5/00 (2019.02); G16B 20/30 (2019.02); G16B 50/00 (2019.02)] | 16 Claims |
|
1. A method for predicting structure information of a protein, performed by a computer device, the method comprising:
performing sequence alignment query in a first database according to an amino acid sequence of the protein to obtain multi-sequence aligned data;
performing feature extraction on the multi-sequence aligned data to obtain an initial sequence feature;
processing the initial sequence feature by using a sequence feature augmentation model to obtain an augmented sequence feature of the protein, the sequence feature augmentation model being a machine learning model trained by using a sample initial sequence feature and a sample augmented sequence feature and updated according to an augmented sample initial sequence feature and the sample augmented sequence feature, the sample initial sequence feature being obtained by performing sequence alignment query in the first database according to a sample amino acid sequence, the sample augmented sequence feature being obtained by performing sequence alignment query in a second database according to the sample amino acid sequence, wherein a data scale of the second database being greater than a data scale of the first database, and the augmented sample initial sequence feature being obtained by processing the sample initial sequence feature by using the sequence feature augmentation model, wherein the sequence feature augmentation model comprises one of:
a fully convolutional network (FCN) model for one-dimensional sequence data;
a recurrent neural network (RNN) model comprising a plurality of layers of long short-term memory (LSTM) units; or
an RNN model comprising bidirectional LSTM units; and
predicting structure information of the protein based on the augmented sequence feature.
|