Page 273 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

264 Chapter 9 Applications of deep learning in biomedical engineering

DNA sequencing: In biomedical data analysis, DNA sequence
classification is the context of generic framework. This helps to
recognize the structural and functional relationship model from
the primary sequence of input data molecules. The deep neural
models such as CNN and RNN are tremendously beneficial to
such kind of tasks. Particularly, RNN handles sequential informa-
tion over time series of data, and it remains an efficient property
in DNA sequence classification [30].
DNA methylation: DNA methylation occurs due to the epige-
netic modification in human genomic DNA. It happens in the
condition of CpG dinucleotide. The DNA methylation patterns
have been used as a biomarker for diagnosing and treating
complex diseases such as cancers. These patterns can be
employed to discriminate the various cancer profiles in DL archi-
tectures [31].

39. Around the protein

In the context of protein modeling, DL can be applied in two
fields:
• Protein Structure Prediction (PSP)
• Protein Interaction Prediction (PIP)
The features involved in various protein prediction problems
are physicochemical properties, protein positionespecific
scoring matrix, solvent accessibility, secondary structure (SS),
protein disorder, and the difference in torsion angles between
real and the predicted one [2].

40. Protein Structure Prediction

PSP is the most challenging topic in structural bioinformatics.
It focuses on extracting information about the structure of pro-
tein folds, which is the chain of amino acids. During the gene
expression process, every protein folds into a unique three-
dimensional structure that determines its specific biological func-
tion. DL frameworks successfully generate representations of
protein folding to analyze its functions. Recent works on protein
structure prediction are as follows:
1. Prediction of 1D structural annotations
2. Contact map prediction
3. Overall structure prediction
4. Building, ranking, and scoring of structural models [32]
The process of protein structure prediction using DL is shown
in Fig. 9.11.