Medchem 410 - Quiz 6 (Lecture 7)

What do -omics look at?

Disease-state vs. healthy state

Data for genome

CNV

LOH

SNP

rare variants

genomic rearrangements

Data for epigenome

miRNA

DNAm

Histone modifications

TF binding

Chromatin accessibility

Data for transcriptome

mRNA

Gene expression

Alternative splicing

non-coding RNA

Data for proteome

Protein expression

Post-translational modification

Cytokine array

Data for metabolome

Small molecules (biomarkers) profiling in blood, serum, urine, CSF, etc.

Methods used for multi-omics data analysis

Feature selection and engineering

Regression-based joint modeling

Matching patterns of eQTL and GWAS

Clustering

Dimensionality reduction

Matrix factorization

Deep neural networks

Data visualization

Applications of multi-omics

Personalized drug prescription and dosing control

Disease detection and classification

Disease survival prediction

Gene regulation discovery

Molecular mechanism discovery

Biomarkers identification

Industrial control of cell cultures

What do multi-omics (data analysis) lead to?

Biological insights

What is the difference between irrational and rational design?

Irrational: design to get more info

Rational: uses what we know about a structure to design

What is the difference between HTS and High Throughput VIRTUAL Screening?

HTS: chemical libraries → tested with whole cells or purified enzymes

High throughput virtual screening: in silico funneling of large libraries to predict hits… can be ligand-based or structure-based

What does ligand-based screening during virtual screenings involve?

Chemical similarity

Pharmacore

QSAR

Machine learning

What does structure-based screening during virtual screenings involve?

Molecular docking

Scoring

How do we determine experimental protein structure and which is the most common?

X-ray crystallography - MOST COMMON

Cryo-electron microscopy (cryo-EM)

Nuclear magnetic resonance (NMR)

Example shown of X-ray crystallography

Alkaline phosphate (ALP) + it’s active cite

How does Cryo-em work?

Freeze fast to see structure → dynamic

Gives near atomic resolution of structure with an electron microscope (Nobel Prize in Chem in 2017)

WHen do we use NMR?

To study protein dynamics

What is Levinthal’s Paradox?

Proteins fold into lowest energy state in seconds → searching the entire landscape of conformations is not possible in this time frame → degrees of freedoms increasing exponentially

paradox suggests that protein folding is not a random search but follows specific pathways and energy landscapes, likely guided by thermodynamic and kinetic principles. The resolution of Levinthal's paradox supports models like the energy funnel hypothesis, which proposes that proteins fold by progressively narrowing down to the most stable structure rather than sampling all possibilities equally.

What are molecular dynamics (MD) used for?

MD is a physics equation used to calculate forces on every atom

What kind of forces does the MD equation look at?

bond

angle

dihedral

improper

van der Waals

electrostatic

timescales of protein motion (0.1-1ps)

domain vibrations (HB vibrations)

timescales of protein motion (1-10ps)

HB breaking rotational relaxation translational diffusion

timescales of protein motion (10-100ps)

collective water dipole relaxation

timescales of protein motion (1-10ns)

sidechain fluctuations

timescales of protein motion (10-100ns)

protein tumbling

timescales of protein motion (us)

conformational transitions

Levels of theory and spatiotemporal scale (top-down)

Continuum theory FEM — Macroscale

Multiphysics, phase field

Coarse-grained MD, MC — Mesoscale

Classical MD

Relax FF MD — Nanoscale

Ab initio DFT — Atomic scale

MD protein folding — accuracy + speed?

Accurate but slow

What does rosetta coarse graining combine?

Physics + data

What is rosetta coarse graining?

In coarse-grained modeling, instead of representing every atom explicitly, the protein is simplified into a lower-resolution model where groups of atoms (e.g., an entire amino acid side chain) are treated as a single unit

Speeds up structure prediction by reducing the computational cost

What is the rosetta de novo folding protocol?

Primary sequence + Secondary structure prediction → generates fragment libraries → Monte carlo fragment insertion → use experimental data and knowledge-based potentials to create energy evaluation of model → low-resolution model → filter to relax backbone (also uses knowledge-based potentials) → final model

What is CASP?

Critical assessment of structure prediction: competition that evaluates the performance of computational methods for protein structure prediction

major role in advancing structural bioinformatics → DeepMind’s AlphaFold 2

What are some remaining challenges when it comes to protein folding?

Dynamics

Point mutations

Protein complexes

Transmembrane proteins

Interactions with non-proteogenic molecules

What solved most of the challenges when it comes to protein folding?

X-ray crystallography but we can’t get structure of protein inside membrane with it

What are cryptid pockets?

binding sites on proteins that are not apparent in the static structure but can open up dynamically under physiological conditions or in response to ligand binding

example: HIV-integrase inhibitor

Overview of molecular docking

Molecular docking → search algorithm + scoring function → docking assessment

Search algorithm of molecular docking

• Systematic

• Molecular dynamics

• Local shape feature matching

• Genetic algorithm

Scoring function of molecular docking

• Force field

• Emperical

• Knowledge-based

Docking assessment of molecular docking

• Docking accuracy

• Enrichment factor

• Prospective pharmacological validation

Benefits + requirements of quantum mechanics

Most accurate

Requires approximating the Schrodinger equation for electronic properties

What does the Rosettafold use to get the All-Atom

Protein sequence

Nucleic Acid sequence

Metal Ion

Small molecule

Covalently modified residue

What does the RoseTTAFold All-Atom consist of

36 main blocks + 4 refinement layers

What can we get using the RoseTTAFold All-Atom

Protein structure

Protein nucleic acid complex

Protein metal complex

Protein small molecule complex

Covalently modified protein structure

What are score-based docking software?

Rosetta

Autodock Vina

Schrodinger Glide

DOCK

MOE

What are ML based docking software?

Diffdock

RoseTTAFold-All-Atom

Alphafold3