The dataset consists of many different types of macromolecules of biological signifiance. The majority of the data records are of proteins. With DNA being the precursor to RNA, which when translated, proteins are the biomolecules that are directly interacting in biological pathways and cycles. Proteins are usually centered around one or a few job which is defined by their family type. For example, we can have a protein that is from a Hydrolase group, which focuses on catalyzing hydrolysis (breaking bonds by adding water) in order to help promote destruction of chains of proteins, or other molecules. Another example would be a protein that is a transport protein, which allows other molecules such as sucrose, fructose, or even water come in and outside of the cell.
With these proteins having different family types, a question arises if it's possible to determine a protein's family type based on sequence. There are notable search engines such as BLAST which has this capability, but it'd be interesting to see if a machine learning approach can do a good job in classifying a protein's family based on the protein sequence.
This is a protein data set retrieved from Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank (PDB).
The PDB archive is a repository of atomic coordinates and other information describing proteins and other important biological macromolecules. Structural biologists use methods such as X-ray crystallography, NMR spectroscopy, and cryo-electron microscopy to determine the location of each atom relative to each other in the molecule. They then deposit this information, which is then annotated and publicly released into the archive by the wwPDB.
The constantly-growing PDB is a reflection of the research that is happening in laboratories across the world. This can make it both exciting and challenging to use the database in research and education. Structures are available for many of the proteins and nucleic acids involved in the central processes of life, so you can go to the PDB archive to find structures for ribosomes, oncogenes, drug targets, and even whole viruses. However, it can be a challenge to find the information that you need, since the PDB archives so many different structures. You will often find multiple structures for a given molecule, or partial structures, or structures that have been modified or inactivated from their native form.
There are two data files. Both are arranged on "structureId" of the protein:
- pdb_data_no_dups.csv contains protein meta data which includes details on protein classification, extraction methods, etc.
- data_seq.csv contains >400,000 protein structure sequences.
Proteins in general can be a type of enzyme, or a signaling protein, structural, and various other choices. A lof of proteins tend to share very similar characteristics, as some proteins are meant to bind in similar regions as others. For example, a Hydrolase enzyme and a Hydrolase inhibitor protein are going to have similar structures as they will target very similar areas. This is reflected in the confusion matrix and heat map. Gene regulator proteins will have a similarity to RNA binding proteins, DNA binding proteins, as well as transcription proteins. The biggest thing to note as well, as the model only uses features of 4 amino acids at most. The possibility of utilizing amino acids of higher degree in theory should be able to create an even higher accuracy.
There is definitely room for improvement for the model. Utilizing factors such as pH, molecular weight, and other components may be able to yield more information on family group. Furthermore, if possible, increase the length of the ngram_range to include more than just 4 characters to allow for higher interaction between the amino acids as reflected in reality