Video summary
The course began with an examination of biomolecular structure, specifically focusing on amino acids, their classification, chirality, and the diverse interactions that govern them. By integrating concepts from physics such as Lennard-Jones potentials, electrostatics, and torsions, the lecture established a framework for determining which molecular states are probable based on free energy rather than just enthalpy. This shift was crucial for understanding the hydrophobic effect and entropy, allowing for a deeper analysis of protein folding and ligand binding as local phase transitions. The ability to interpret structural stability through these thermodynamic lenses enabled surprisingly accurate predictions using simple calculations, which could be significantly enhanced by computer simulations involving large systems with realistic environments like water.
As the discussion progressed, the focus expanded from individual states to the kinetics of how structures form and transition between different configurations. While protein folding served as a primary example, the principles of kinetics applied universally to other processes such as ligand binding at specific sites. The lecture also highlighted the role of bioinformatics, which analyzes sequence data rather than direct structural information. This approach is effective because four billion years of evolution has encoded physical interaction rules into protein sequences, allowing researchers to infer structural properties and stability directly from genetic data. This evolutionary context explains why certain folds are structurally stable and have been selected over time, bridging the gap between computational analysis and biological reality.
In the final stages of the course, these various disciplines—physical knowledge of protein structure, bioinformatics, and simulation techniques—were combined to advance drug discovery through methods like molecular docking. Docking acts as a powerful tool for predicting binding energies and exploring partition functions without the need for extensive simulations, making it widely used in both academic research and pharmaceutical industries. The curriculum concluded with an exploration of protein design, where scientists explicitly incorporate amino acid properties and Boltzmann distributions to explore regions of the energy landscape that evolution has not yet visited. This forward-looking approach represents a synthesis of theoretical physics, biological data, and computational power, offering new ways to engineer molecules and understand complex biological systems beyond what nature has currently produced.
Read the full video transcript
so we started out a few weeks ago
studying biomolecular structure and in
particular the amino acids why they have
their properties they have what amino
acids how we classify them
and how they give rise to the diversity
and also the chirality that i touched
upon today with d and l amino acids
from there we went on to looking at the
molecular structure and in particular
all the interactions lennard jones
electrostatics torsions
everything we then went back to physics
because based on those interactions we
needed tools that enable to say
what states are probable what states are
less probable
that then led to the hydrophobic effect
where we particularly introduced entropy
right and that i was able to go from
there to talk about free energy instead
of just
enthalpy which is an exceptionally
important concept that kept coming back
to the class
we went on to instead of just starting
pairs of states studying the entire
partition function
all possible states and from that we
could start to talk about things in
terms of stability of different
states sure i used water as the primary
example here but this is equally true
for protein folding as i showed you a
few lectures ago
and even things like ligand binding you
can think of this as a very small
local phase transition for that ligand
we
used that to interpret protein
structures in particular small parts of
the structure
in terms of free energy rather than just
entropy that
enabled us to explain roughly how fast
they form and everything which is so in
hand in task too that we can make
surprisingly good predictions based just
on paper and pen
although if we combine that with a
computer we're able to study
much larger systems where we can have a
realistic environment
water tens of thousands of atoms and
make actual predictions about
actual systems that we can compare to
experiments not only that we could use
so-called free energy calculation
methods to make specific predictions
about
how good should it be for a molecule to
bind in a specific place
in the molecule or in your case what to
the probability of the binding energy
for c non b
in a particular site we
looked quite a lot about protein
structure first the fibrous and the
water soluble proteins
then the membrane proteins and also the
lipids that surround the membrane
proteins
we focused even more than on what are
these structures what do we have them we
group them in folds we started to reason
why the faults we see are structurally
stable
and why evolution has selected for those
faults we're going to come back to the
evolution
and a few lines here we then went on to
study not just the states the faults we
have but
how they form that is the kinetics when
we're moving from one state to another
now we primarily used folding as the
simple universal examples here but as i
showed you today this was equally true
for the ligand binding to some of those
binding sites
so don't think that this is a narrow
example limited to research and protein
folding it's not
what you learned about the kinetics is
universally true
based on the importance of the evolution
here though in lecture 11 we focused on
bioinformatics which on the one hand is
a completely different approach just
focusing on
on sequence and data instead of
structure but the reason why that works
so well
is that four billion years of evolution
has encoded all the physical
interactions here
in the protein structures we see and
that means that i can compare the
sequence to effectively compare the
structures
it's really cool that it works and the
only reason why that does work is
evolution last lecture
we then combined both the physical
knowledge the protein structure the
bioinformatics and a bit of the
simulation
to come up with drug discovery and
particularly using docking effectively
as
poor man's way of doing a simulation or
studying the partition function
it is a remarkably powerful tool that is
used throughout both academia
and pharmaceutical companies today and
the last hour
we have now finished up by trying to
combine that with
simulations and particularly how we can
use protein design
where we now even more explicitly try to
include the amino acid properties and
the boltzmann distribution
to go to the places of the map where
evolution has not
yet been it's been a
great pleasure for me to take you
through this class
lots of work preparing these recordings
but it's very fun
so let me thank you for watching it this
far and in particular if you're in
stockholm i hope to get a chance to see
in the future
thanks