DeeperBind: Enhancing Prediction of Sequence Specificities of DNA Binding Proteins

Abstract

Transcription factors (TFs) are macromolecules that bind to cis-regulatory specific sub-regions of DNA promoters and initiate transcription. Finding the exact location of these binding sites (aka motifs) is important in a variety of domains such as drug design and development. To address this need, several in vivo and in vitro techniques have been developed so far that try to characterize and predict the binding specificity of a protein to different DNA loci. The major problem with these techniques is that they are not accurate enough in prediction of the binding affinity and characterization of the corresponding motifs. As a result, downstream analysis is required to uncover the locations where proteins of interest bind. Here, we propose DeeperBind, a long short term recurrent convolutional network for prediction of protein binding specificities with respect to DNA probes. DeeperBind can model the positional dynamics of probe sequences and hence reckons with the contributions made by individual sub-regions in DNA sequences, in an effective way. Moreover, it can be trained and tested on datasets containing varying-length sequences. We apply our pipeline to the datasets derived from protein binding microarrays (PBMs), an in-vitro high-throughput technology for quantification of protein-DNA binding preferences, and present promising results. To the best of our knowledge, this is the most accurate pipeline that can predict binding specificities of DNA sequences from the data produced by high-throughput technologies through utilization of the power of deep learning for feature generation and positional dynamics modeling.

Fig. 1:

Recurrent neural networks. (a) Diagram of an LSTM memory cell used in this paper. Small pink circles denote pointwise operations, arrows indicate vector transfer, merged arrows mean concatenation of vectors and split arrows indicate vector copying operation. (b) concatenation of a few LSTM cells to build a deep LSTM network.

Fig. 2:

Block diagram of DeeperBind. The input sequences are first represented as 2D binary matrices via one-hot coding. A convolutional layer generates the feature map by applying several PWM-like filters followed by rectified linear units. No pooling layer is used. Two stacks of LSTM layers then capture the sequential dependencies of the sub-motifs on probes.

Fig. 3:

The predicted rank of the top 100 positive probes (black lines) in array #1 for the DeepBind (left) and the DeeperBind (right) per each TF: a) CEH-22 and b) Oct-1.

Fig. 4:

Scatter plot of predicted and measured intensities: DeepBind vs DeeperBind. a) CEH-22 and b) Oct-1.

Fig. 4:

Scatter plot of predicted and measured intensities: DeepBind vs DeeperBind. a) CEH-22 and b) Oct-1.

Fig. 5:

Receiver Operating Characteristic (ROC) curves on array #2. Samples are marked as positive if their measured intensity exceed m + 4σ where m is the and σ is the median absolute deviation of all probe normalized intensities divided by 0.6745.