Biotechnology: Class 11 — Chapter 10: Protein Informatics and Cheminformatics

10.1 Protein Informatics

10.1.1 Introduction & Protein Data Types

Collecting information about any protein using techniques of information technology comes under protein informatics. It helps in identifying the geometrical location of functional sites, biochemical/biological functions of hypothetical proteins, and tertiary structures of proteins whose functions cannot be determined by conventional methods.

Raw protein data types include:

Microscopic image of heat-denatured protein aggregate (used for designing protein-markers via multi-fractal properties).
Protein in solution form (useful for analyzing physico-chemical properties and kinetics).
MALDI output (Matrix Assisted Laser Desorption Ionisation) fragmented short sequences used to find full-length sequences.
Assembled protein sequence.
Protein crystal structure in Protein Data Bank (PDB) format (used to study mutations and interactions).
Interaction files (protein-protein, protein-ligand, or protein-nucleotide).
NMR data (Nuclear Magnetic Resonance) and Mass Spectrometry (MS) data (used for structure prediction of non-crystallized proteins).
Hypothetical proteins derived directly from genomic sequences without prior evidence of physical existence.

Two basic facilities required for protein informatics analysis:

Availability of raw data from databases like NCBI, PDB, ChEMBL, and BIOMODELS.
Informatics tools and techniques: Wavelet image analysis, sequence similarity/homology calculations, structure optimization, machine learning (ANN, SVM, HMM), Network Mapping, and SBML (Systems Biology Mark-up Language).

10.1.2 Computational Prediction of Protein Structures

Aimed at exploring how amino acid sequences specify protein structures and binding properties. Major advantages include speed, low cost, and high-throughput screening feasibility.

Primary Structure Prediction

Involves physico-chemical characterization calculated using the ProtParam tool of the ExPASy Proteomics Server:

Isoelectric point (pI): The pH at which the net charge of the protein is zero. Proteins are stable and compact at pI. If $pI < 7$, the protein is acidic; if $pI > 7$, it is basic. Useful for buffer system design in isoelectric focusing.
Aliphatic index (AI): Relative volume occupied by aliphatic side chains (A, V, I, L). High AI indicates high thermal stability over a wide temperature range.
Instability index: Estimate of stability in a test tube. Weight values are assigned to dipeptides. A value $< 40$ predicts a stable protein; $> 40$ predicts instability.
Grand Average Hydropathy (GRAVY): Sum of hydropathy values of all amino acids divided by the number of residues. Low GRAVY indicates better water interaction.

Secondary & Tertiary Structure Prediction

Secondary structure prediction tools include APSSP, CFSSP, SOPMA, and GOR.

Three computational methods for 3D structure prediction:

Homology modelling: Aligns target sequence against known structures. High homology determines global structure. Tools: MODELLER, SWISS-MODEL.
Fold prediction (Threading): Forces target sequence to adopt the backbone conformation of a known structure. More compute-intensive. Tools: LIBELLULA, Threader.
De novo protein structure prediction: Algorithmic prediction of tertiary structure directly from primary sequence. Tool: QUARK (for ab initio prediction and peptide folding).

Elucidated structures are stored as atomic coordinates in PDB files (extension .pdb) containing X-ray crystallography, NMR, and theoretical model data.

Domain prediction: Identifies distinct functional/structural units of a protein. Tools: InterPRO scan of EMBL, CDD search of NCBI.

10.2 Cheminformatics

10.2.1 Introduction & Chemical Databases

The use of computational and informational techniques to understand problems of chemistry is known as cheminformatics. It combines physics, chemistry, biology, mathematics, biochemistry, statistics, and informatics to aid in drug discovery and virtual screening.

The CAS (Chemical Abstracts Service) registry is the world's largest collection of chemistry insights, hosting over 219 million substances, 70 million sequences, and 8 billion property values.

Database Name	Description
PubChem	Maintains substance, compound, and BioAssay information.
ZINC	Contains compounds for virtual screening with molecular weight, log P, etc.
ChEMBL	Comprehensive database of bioactive small drug-like molecules with targets.
NCI	Small molecule structures database, highly useful for cancer/AIDS research.
ChemDB	Chemicals database with predicted/experimental properties (3D structure, melting point, solubility).
ChemSpider	Aggregated unique chemical entities from diverse data sources.
BindingDB	Binding affinity database of small molecules for protein targets.
DrugBank	Combines detailed drug data with comprehensive target, sequence, and pathway info.
PharmaGKB	Pharmacogenomics knowledge resource with clinical drug information.
SuperDrug	Contains 3D structures of active ingredients of marketed drugs.

10.2.2 Representation, Searching & Pharmacophores

Chemical structures are stored in computers as molecular graphs, where atoms are represented as nodes and bonds as edges. This node-edge approach is also used to map metabolic pathways like glycolysis and the Krebs cycle.

Structure searching methods:

Substructure retrieval: Finds chemical compounds containing specific functional groups (e.g., methyl group, benzene ring). Uses subgraph isomorphism.
Two-stage search: First filters out non-matching molecules using binary strings (bitstrings of 0s and 1s), then performs elaborate subgraph isomerism on the remaining candidates.
Reaction searching: Searches reaction databases for products, conditions, and pathways. Uses atom mapping to find exact correspondences between reactant and product atoms.

A pharmacophore is an ensemble of steric and electronic features necessary to ensure optimal interactions with a specific biological target to trigger a biological response (IUPAC definition). It is a conceptual framework, not a physical molecule.

10.2.3 Lipinski's Rule of Five (RO5)

Proposed by Christopher A. Lipinski in 1997, the RO5 describes key molecular properties for predicting if a compound is likely to be an orally active drug. An orally active drug should have no more than one violation of the following criteria:

Rule Parameter	Threshold Limit
Hydrogen Bond Donors	Not more than 5
Hydrogen Bond Acceptors	Not more than 10
Molecular Weight	Below 500 Daltons
Octanol-Water Partition Coefficient ($\log P$)	Less than 5

Based on RO5, compounds are assigned a score of 0-4. A score less than 3 indicates the compound is not suitable for further analysis. Note: RO5 does not apply to natural products, semi-synthetic natural products, or drugs administered via intramuscular/intravenous routes.

10.2.4 Drug Discovery Pipeline & Virtual Screening

The journey of a drug from lab to market is long, expensive, and risky. The pipeline is structured as follows:

Phase	Duration	Compound Count	Key Steps / Milestones
1. Discovery	5–6 years	Thousands of Compounds	Idea generation, basic process research, virtual screening.
2. Development	7–8 years	1–5 Compounds	Clinical trials (Phase I, Phase II, Phase III).
3. Delivery	1–5 years	1 Compound	Regulatory approval, patient care.

Virtual screening is an in-silico approach to score, rank, and filter compounds. It utilizes:

General filters for ADME (Absorption, Distribution, Metabolism, and Excretion) properties.
Ligand-based methods (machine learning, pharmacophore search).
Structure-based methods (protein-ligand docking).

Box 1: Accidental Discoveries
Viagra (UK92480): Originally tested by Pfizer in the early 1990s to relax blood vessels and treat heart complications; unexpected side effects on the reproductive system led to its development as a treatment for erectile dysfunction.
Saccharin: Discovered accidentally in 1879 by Russian chemist Dr. Constantin Fahlberg, who forgot to wash his hands before dinner after working with coal tar derivatives and noticed an intense sweet taste on his bread.

Box 2: Common Terminologies
High Throughput Screening (HTS): Large-scale automated testing of millions of compounds.
Hits: Observed activity during screening compared to known compounds.
False positive: Active in assay but inactive against the actual biological target.
Lead compound: Biologically active compound with desired properties selected for further optimization.
Library: Inventory of compounds for target screening.
New Chemical Entity (NCE): Novel molecule discovered in the lab not yet in clinical trials.
Off target activity: Unintended molecular interactions with non-target cellular molecules.