But it was not completely true. Many aspects of the functions and protein modulation could not be explained by merely looking at the static molecular machine. Much the same way, we cannot understand anything about the power, speed and fuel efficiency of a car by merely staring at it parked in a showroom.

Modern biophysics has given us a rather different picture. Proteins are not static molecular machines. They are dynamic, continuously fluctuating among multiple conformational states, with each state populated according to its free energy. At a given temperature, the Boltzmann distribution determines how these states are populated. What we call “the structure of a protein” is therefore often nothing more than the most probable snapshot of a much larger and constantly changing conformational landscape.

This becomes particularly interesting when we think about molecular recognition.

Binding is not simply a matter of a rigid ligand finding a perfectly complementary lock. Both the receptor and ligand occupy ensembles of conformations, and their interaction is governed by a competition between these states, their populations, and the energetic and entropic consequences of bringing them together. A ligand may preferentially bind to a pre-existing receptor conformation, or binding itself may shift the population toward a new conformational state. In either case, molecular recognition is better understood as a process occurring on a free-energy landscape than as a simple lock-and-key event.

The more I read about molecular simulations and statistical mechanics, the more I found that some of the most interesting insights are also the most counterintuitive.

Entropy is essential, but often skipped

Entropy is often portrayed as the villain, the measure of disorder, randomness, and the inevitable tendency of things to fall apart. In thermodynamics, however, entropy is far more interesting than that. In the molecular world, entropy is not simply chaos. It is a measure of the number of microscopic states available to a system (think of it like what standard deviation does to mean). And in protein evolution, dynamics, and molecular recognition, that multiplicity of states can become a functional advantage.

One of the fascinating consequences is what may be described as entropic rearrangement. When a protein binds a ligand, the binding site may become more ordered. Locally, the protein loses conformational freedom. But thermodynamics does not require the entropy change to be confined to the binding site. Other regions of the protein can respond by becoming more flexible, effectively redistributing conformational entropy across the molecule.

In other words, the protein does not necessarily become “more ordered” as a whole. It can simply move the disorder somewhere else.

An example is found in PDB ID 9NT5, where binding of a partner is associated with the unravelling of a distal α-helix. At first glance, this structural disorder may appear to be an unwanted consequence of binding. Thermodynamically, however, it can represent part of the energetic and entropic compensation that makes the overall process favourable. This is one reason I find entropy such an interesting concept. What appears to us as disorder at one location can be an essential component of order at the level of the entire system.

As Colin J. Jackson has argued, entropy can be viewed not simply as a force pulling systems toward chaos, but as a selective pressure that can shape successful evolution. In sufficiently complex environments, flexibility and the ability to redistribute conformational states may itself become a form of molecular adaptability.

Noise can improve a simulation

There is another counterintuitive lesson in molecular dynamics.

If we wanted to simulate a physical system perfectly, our first instinct might be to make the simulation as deterministic as possible. After all, classical mechanics is deterministic: if we know the positions and velocities of every particle, Newton’s equations tell us how the system evolves.

But statistical mechanics introduces a subtle problem. A deterministic trajectory is not necessarily guaranteed to explore the entire phase space. A thermostat such as Nosé–Hoover can generate the desired statistical ensemble under appropriate conditions, but in difficult systems such as harmonic solids, the dynamics can become non-ergodic. The trajectory may remain confined to only a small subset of the states that should, in principle, be sampled.

And here comes the irony. The solution to this problem can be to add randomness. Stochastic velocity rescaling, introduced by Bussi and colleagues, deliberately introduces a stochastic component into the dynamics while maintaining the desired canonical distribution. Randomness, which at first appears to be an imperfection, can therefore improve the ability of a simulation to sample the correct statistical ensemble. This is one of those ideas that feels almost paradoxical at first. If the physical system is deterministic at the microscopic level, why should adding randomness make the simulation more physically meaningful?

The answer lies in the distinction between microscopic trajectories and statistical ensembles. We are not necessarily interested in reproducing one exact trajectory. We are interested in sampling the correct distribution of states.

The purpose of molecular dynamics is therefore not always to eliminate randomness. Sometimes it is to introduce precisely the right amount of it.

As Bussi and co-workers noted, deterministic schemes can suffer from poor ergodicity and inadequate temperature control in difficult systems. Stochastic dynamics provides a way around this limitation. The apparent noise is not the problem. Sometimes, it is the solution.

Breaking the “memoryless” assumption

Another assumption that becomes increasingly uncomfortable as we move toward complex biological processes is the idea that molecular dynamics can always be treated as Markovian. In a Markovian description, the future evolution of a system depends only on its present state and not on the details of how it arrived there. This approximation is extremely useful. But biological processes often span a hierarchy of time scales, and the system may retain a memory of its recent history.

Protein folding is an obvious example. A molecule may spend a considerable amount of time exploring local conformations before making a transition to another state. The probability and timing of that transition can therefore depend on the history of the system. This is where the work of Faradjian and Elber on milestoning becomes particularly interesting. Instead of waiting for a conventional simulation to sample an entire transition from one metastable state to another, one can place a series of hypersurfaces or milestones along a suitable reaction coordinate. The simulation then focuses on the first-passage events between these milestones. The important information is not merely whether the molecule reaches the next milestone, but how long it takes to get there and how the system approaches the transition. These first-passage-time distributions retain information about the underlying dynamics that can be lost in an overly simplistic Markovian description. There is an interesting balance here. The milestones must be sufficiently separated that the system has time to lose memory of its previous transition, but sufficiently close that the calculation remains computationally tractable. The result is a framework that allows us to extract long-time kinetic information from simulations that would otherwise be prohibitively expensive.

As Faradjian and Elber pointed out, the non-Markovian description is more general and provides additional flexibility in modelling molecular processes. Once again, the lesson is that a convenient assumption can become a limitation when the system itself refuses to behave conveniently.

The arbitrariness of classical force fields

Classical force fields (FF) are another reminder of how much approximation is hidden beneath the apparent precision of molecular simulations. At their core, these models are simplified (often over-simplified) representations of molecular physics. They describe bonded interactions, electrostatics, van der Waals interactions, and other contributions using parameterised functional forms. They are remarkably useful, but they are not fundamental theories of chemistry. This becomes particularly obvious when we examine 1–4 interactions between atoms separated by three covalent bonds. In several force field families, these interactions are treated differently from ordinary non-bonded interactions. Scaling factors are introduced to modify the van der Waals and electrostatic contributions. In the AMBER FF, for example, the conventional values correspond to f_vdW = 0.5 and f_e = 1/1.2, commonly represented in software using parameters such as scnb = 2.0 and scee = 1.2. The OPLS FF takes a different approach, while CHARMM generally does not apply the same 1–4 scaling scheme and instead uses a different treatment of these interactions. These numbers can look strangely arbitrary when encountered for the first time.

Why 0.5? Why 1/1.2? Why not another value?

Because a force field is ultimately a carefully calibrated approximation. These parameters are not universal constants derived from first principles. They are part of a model designed to reproduce selected experimental and quantum-mechanical properties with a computationally tractable functional form. Calling them “hacks” is perhaps deliberately provocative, but there is a useful truth hidden in the provocation. Computational chemistry is full of such compromises. We construct a model that is sufficiently simple to simulate millions of atoms and sufficiently accurate to reproduce the phenomena we care about. The important question is therefore not whether a force field is an exact representation of reality. It never was and never will be for a foreseeable future. The important question is: for which physical questions is the approximation reliable?

Measuring an effective enthalpy in a stochastic universe

There is another problem that appears when we introduce stochastic dynamics. In a conventional microcanonical simulation, one of the simplest ways to assess numerical stability is to monitor the total energy. If the integrator is behaving properly, the total energy should remain approximately conserved, with only small fluctuations or a controlled numerical drift.

But what happens when the simulation itself contains random forces?

With stochastic thermostats and barostats, particularly in simulations designed to sample the NPT ensemble, the ordinary Hamiltonian is no longer conserved. The random terms are deliberately adding and removing energy. Does that mean we have lost one of our most useful diagnostics? Not really.

For certain stochastic dynamical schemes, one can construct an extended or effective conserved quantity, often referred to as an effective enthalpy, H_eff, that incorporates the contributions associated with the stochastic dynamics. Tracing its behaviour provides a way to distinguish statistical fluctuations from systematic numerical errors. This is conceptually important. The disappearance of an ordinary conservation law does not mean that the simulation has become impossible to validate. It means that we need to understand what quantity should be conserved, or approximately conserved, under the particular equations of motion we have chosen. A persistent drift in an appropriate effective quantity can reveal problems with numerical integration or inadequate simulation parameters. In some model systems, reducing the integration timestep can be necessary to control such errors. The broader lesson is that every simulation comes with its own definition of what “correct” means. A thermostat changes the equations. A barostat changes the equations. A stochastic integrator changes the equations. And therefore, the diagnostics must change with them.

The future is ensembles

Perhaps the most important conceptual shift in molecular simulation is not a new thermostat, a new force field, or a new sampling algorithm. It is the gradual disappearance of the idea that a molecule can be understood from a single structure. We have traditionally looked at proteins as structures and then tried to explain their function from those structures. Molecular simulations increasingly forced us to change this view starting from the first MD simulation by Martin Karplus. A structure is only one point in a much larger conformational landscape. Function emerges from how the system moves through that landscape, how states are populated, how transitions occur, and how interactions reshape those populations. The molecule is not the structure but the ensemble. And once we accept that, many of the apparent contradictions in molecular simulation begin to make sense. We introduce scaling factors because our force fields are approximations. We introduce stochasticity because deterministic dynamics can fail to sample the desired ensemble. We develop non-Markovian methods because biological systems retain memory across multiple time scales. We monitor effective conserved quantities because conventional energy conservation no longer applies to stochastic dynamics. At first glance, these methods can look like a collection of computational tricks. Indeed, they ought to be. But there is another way to look at them. They are the tools we have developed because biological molecules are themselves complicated, fluctuating, probabilistic systems. The “hacks” are not signs that our understanding has failed. They are reminders that no single mathematical description captures every level of molecular reality. For a long time, we tried to eliminate the chaos from our picture of molecular biology. We are now beginning to understand that the chaos was part of the picture all along. Perhaps the future of molecular simulation will not be about finding a perfectly deterministic description of life. Perhaps it will be about learning how to describe, quantify, and ultimately exploit the fluctuations that make life possible.

At the atomic scale, what we once regarded as noise may turn out to be information.

Membrane permeability is a crucial step in understanding the bioavailability of a drug, which in turn suggests how efficiently drug molecules can reach the site of action.
Classical computer simulations generally fail to sample the high barrier regions that the drug must surmount to permeate the bilayer membrane.
In this video, we show how steered molecular dynamics and umbrella sampling can help estimate the membrane permeability of drug like molecules.

Please subscribe to my YouTube Channel: The MD Guy

Classical MD methods often fail to sample the regions on the phase-space that are separated by high energy barriers. In practice, any barrier large than 1 KbT poses significant sampling issues for classical MD methods. To solve this challenge, researcher have learned and developed methods that artificially tweak the potential energy surface of the system so that the system can surmount the barriers.

In this video, we show few such methods and their application in drug discovery.

Please Subscribe to my YouTube Channel: The MD Guy

Please subscribe to my Youtube Channel: The MD Guy

In this video, we are discussing yet another trick in the trade, coarse-graining. This helps to increase the time-length of atom events, however, it loses the atomic details.

We explain how such models are built and when is it appropriate to use coarse-grained models.

I hope you like this video.

Please subscribe to my Youtube channel: The MD Guy

GET YOUR FUNDAMENTALS RIGHT
KNOW THE FUNCTION OF EVERY BUTTON YOU PRESS OR COMMAND YOU ENTER.

This was enough for chills to run down my spine. I had no idea what those buttons in Desmond did. Someone helped me setup and run the simulations, I was exhilarated looking the movie, but I had no idea what biology or chemistry the movie was supposed to narrate. To my horror, I didn’t know why I was doing it. I had read a few papers (actually a little more than few), and noted the flow of methods. That was it, I planned my work in the same way. And since most, virtual screening and hit identification did use MD simulations, so did I.

I silently waited for my turn while watching other before me getting bombarded with questions. Everyone but a few PhDs struggled to answer all the questions.
It was my turn. I quickly loaded my slides, and jumped to show and discuss the new movies and plots. I was just talking what I did. And as expected, the question:

Why did you do this?
What do you intend to achieve from this?

By now, you might have figured out that I had no answer. 10 minutes of my presentations, and 60 minutes of how NOT to do science talk for everyone. Since many before me also failed to give satisfactory answers: this talk was much needed. Even though most of us knew what was being said, it was out sheer laziness and impatience to get some answer/output that we ignored the important things. The fundamental concept in research:

Trust but Verify

From there on, I tried to understand everything. I swallowed the Desmond manual. In the end I knew what every button and setting was supposed to do, and its effect on the outcome. One thing was still missing, I was clueless about the force field parameters it was using, and how it managed to get parameters for ligand molecules, which are generally unique for every research problem. Desmond is a plug and play tool, wherein, the MD engine was bundled with a powerful graphical user interface called “Maestro” developed and supported by Schrödinger Inc. It was easy to load the protein (also other systems), follow the protein preparation wizard, and soak the system in water and ions using Desmond’s system preparation wizard. Once you are satisfied with everything, just open Desmond’s MD tool and enter the system parameters, such as length of MD run (in nanoseconds), decide what kind of equilibration protocol one wants to follow, decide the ensemble for production (NVT or NPT), and decide how frequently one wants to save conformations for analysis. Then press the run button with CPU or GPU to be used for computations. There you go, within a few hours or days, depending on how powerful the computing system is, you would get the output files. Unless of course, something went wrong and the calculations died or failed. There could be several reasons.

One can see that the workflow is pretty much straightforward and little to no command-line knowledge is required, unless you prepare the system on a local machine and wish to submit the job on a supercomputer far away.

Following the days after the horrible lab meeting presentation, I was trying to understand every bit of information about MD simulations. Seeing this, my advisor suggested that I learn tools like AMBER or GROMACS, which allowed many types of calculations. I would learn them one-by-one over the years (still learning 12 years after I started as a beginner, and still would call myself an expert beginner!!). After some talking around in the Lab, I narrowed down to learning the AMBER suite of tools. One of the reasons, it was already available in the lab and someone was already using. Unfortunately, the one person with this know-how left the lab for better prospects, and I was left to fend for myself. Nevertheless, since I had decided to learn AMBER, there no looking back (Surely, there was but I decided not to).

AMBER had a very different philosophy of working. Unlike Desmond, every parameter was to be provided by the user. More importantly, it was a complete command-line based tool, no GUI except a few progams. For biomolecules, it was straightforward, however, when you were working with small drug-like molecules, you had to extract parameters using semi-empirical or Quantum Mechanic calculations. At the beginning, I followed an excellently writing AMBER tutorial page. This page is being constantly updated as new and easier ways of doing things are reported. This page was my go-to site when I was struggling to understand how to do something. There were of course 100s of errors when I started off, AMBER mailing list, is yet another community-driven place where people reported bugs and asked help on issues not directly mentioned in the user manual. It took me nearly a year to understand the AMBER way of working with systems.

During November of 2014, my advisor got an invite to an Indo-German winter school and conference in Heidelberg. He was allowed to get one student along and it would be fully paid trip jointly supported by India’s Department of Science & Technology (DST) and Germany’s Deutsche Forschungsgemeinschaft (DFG). To my pleasant surprise, the Indian students would be supported to stay back for an 8-week exchange program with a select German host professor. I was extremely fortunate to have gotten CPCLAB in Düsseldorf headed by Prof Dr Holger Golhke. He was one of the developers for AMBER MD tool. There was nothing more I could have asked for. I had landed in a small Mecca of Molecular dynamics simulations. During this, I learnt how to run MD simulations that could be tested statistically (Replicate MD runs, which I have noticed most ignore). How to use quantum mechanical calculations to derive small molecule parameters. And how to use MMPBSA and MMGBSA as postMD methods to estimate the binding affinity of small molecules.

These methods were central to my PhD thesis, and I have used them whenever needed. After my PhD, I also tried and automated the binding affinity workflow. It could efficiently prepare protein-ligand systems following a virtual screening program for more than 100,000 molecules to rescore before selecting good ones for experimental testing. This workflow also made my life easy during my postdoc projects.

There were several occasions, I would have loved to have some Youtube videos on this theme. There were few, however, those would only explain the most basic topics in MD and not touch those areas that were truly fundamental to MD.

I have to tried to address these gaps. I agree that I am still far from making high quality content. But this is my first attempt at using Youtube for disseminating my teaching content through my Youtube channel “The MD Guy.”

Few a weeks ago, I finished uploading all the contents meant for beginners under the playlist “Molecular Dynamics 101”

Here is the list of videos from that playlist. I hope you enjoy learning.
Please send in your comments, feedbacks and suggestion in the comment section.

• First Video: Molecular Dynamics 101: Basic Introduction to MD simulations
  -

• Second Video: Molecular Dynamics 101: Importance of Force Fields

• Third Video: Mechanics of Potential Energy Surface

• Fourth Video: The Ergodic Hypothesis

• Fifth Video: Want to know how exactly MD simulations works?

• Sixth Video: Tricks of the Trade in MD simulations

• Seven Video: The heart of MD simulations: Non-bonded Interactions

• Eight Video: The Geometry of Space: Periodic Boundary Conditions

• Ninth Video: Approaching the practical aspects of MD Simulations

• Tenth Video: Know how to validate your MD simulations

I hope you enjoyed this video.

Please subscribe to my Youtube channel: The MD Guy

Several different models can replicate different properties that match the experimental values, however, not a single models can replicate all the properties of water.

These ones that can repicate good lot of bulk properties of water are extremely slow in terms of simulation speed but most accurate. However, when we try to enhance the speed, we start to lose accuracy.

This video will not help you resolve this dilemma, however, it will show where the problems lies, and what researchers are doing to solve it. This also explains why several different water models exist for biomolecular simulations.

I hope you will like this video, as much as I liked making it.

Please subscriber to my YouTube Channel: The MD Guy.

In this video, we explain the tricks and tips to prepare input files and the checklist the ensures the MD results are valid.

The checklist will prepare you such that you are ensure corretness of input preparation to valid MD trajectories.

This is the last video in the molecular dynamics 101: theory series. I hope I have covered everything that one needs to get a grasp of MD simulations.

I hope you liked this video and the series as well

Thank you for watching.

Please subscribe to my YouTube Channel for more interesting concepts: EMs_MDLabs.

From this video onwards, we start to understand more practical aspects of MD simulations.
We attempt to illustrate the following steps/concepts that need to be understood to yield a more reliable output from computer simulations:

1) The source of experimental structures as input for computer simulations.

2) What are points that need to be kept in mind while preparing the system.

3) Why are counterions important in the solvent box of MD simulations?

4) How to estimate the number of ions to maintain the 0.15 M concentration that mimics physiological salt concentration?

5) Why does the system’s charge should be strictly neutral to employ the particle mesh Ewald sum method effectively?

I hope you liked this video.

Please subscribe to my YouTube Channel: @EMs_MDLabs

In molecular dynamics simulations, we can only simulate a tiny box containing a limited number of atoms or molecules. However, real materials and biological systems are effectively infinite in size. To overcome this limitation, scientists use Periodic Boundary Conditions (PBC).
Under PBC, the simulation box is imagined to repeat itself endlessly in all directions, like tiles in a 3D wallpaper. When a particle leaves one side of the box, it immediately re-enters from the opposite side with the same velocity. This creates the illusion of an infinite system and prevents artificial surface or edge effects.
PBC allows small simulations to behave more like real bulk systems, improves statistical accuracy, and enables realistic modeling of liquids, proteins, membranes, and materials while keeping computational cost manageable.

In short, Periodic Boundary Conditions make a small simulation box behave like a tiny window into an infinite universe of molecules.

I hope you liked this videos.

Please subscribe to my YouTube Channel: EMs_MDLabs

The core of molecular dynamics (MD) simulations lies in the non-bonded interactions that likely dictate the dynamics of the particles within the simulation box. Calculating these interactions proves intricate due to the extensive mathematical operations required to account for all particle-particle interactions. However, this process would necessitate an infinite amount of time and computational resources.

In this video, we present another set of “tricks of the trade” in MD simulations that enhance computing speed while preserving the physical reality of the biological system.

I hope you liked this video.

Please Subscribe to my YouTube Channel: @EMs_MDLabs

While doing so I tried my hand at NotebookLM. Here I would search and add material as source and try to ask relevant question to notebookLM. It was interesting. Then, I started making video illustrations to make good (not best yet) explanation with visuals. And sooner than I thought, these videos landed (I put them there) up on my YouTube Channels (Made only for this). Instead of writing blogs, I was trying to master how to write good prompts to get better video illustrations.

So I landed up making a YouTube channel. It talks about the field I love the most and have been working and trying to get better at it; Molecular Dynamics Simulations for Biomolecular Systems. I used and read various books and material from the internet. Then I fed these material to notebookLM and started making videos Illustrations.

The sole purpose is to learn things, and also to share what I learn with everyone.

Thank you for reading this.

I hope you will like it.

I will be back with more such content soon.

Here are the link to my Youtube Channel:

EMs_MDLabs

1. Latest Video

The sixth video in the Molecular Dynamics 101 series we cover concepts that must be understood before embarking on setting up MD simulatins. This video covers the following important questions:

• What are the differente thermodynamic ensembles that commonly used in Molecular dynamics simulations?
  

• What are thermostats in MD simulations?
  

• What is basic concept behind thermostats in MD simulations?
  

• Different types of thermostats and their how they work to control the temperature of the system?
  

• What are barostats in MD simulations?
  

• What is basic concept behind barostats in MD simulations?
  

• Different types of barostats and their how they work to control the pressure of the system?
  

• Why must one choose constant volume ensemble for heating the system in MD simulations?
  

• Why is it isobaric-isothermal ensemble efficient in density equlibrations?
  

• Why simulations understand constant volume ensemble are computationally faster than simulations under isobaric-isothermal ensemble?
  

• What is an optimal choice thermodynamic simulation to closely mimic biological systems?
  

I hope you enjoyed this video.

Please subscribe to my YouTube Channel: @EMs_MDLabs

In this fifth video of the series, we dwell deep in to the working of the MD simulations.

How the motions of integrations are solved to move the atom?

How velocities are first assigned to the atoms based on a target temperature?

I hope you liked this video.

Please subscribe to my Youtube channel: @EMs_MDLabs

“Most molecular docking workflows look rigorous — until you try to reproduce them."

This is something I kept encountering while working in computational drug discovery, as well as students frequently asking similar doubts during teaching sessions and workshops.
Docking is often presented as a clean, structured pipeline, however, in practice it is highly sensitive to input preparation, receptor quality, scoring assumptions, and interpretation bias.
Yet, we continue to treat docking outputs as if they are intrinsically reliable and absolute truth. This disconnect is precisely what motivated me to explore the field more deeply and started to write about my experiences that I learnt during my projects and teaching classes on this subject.

I’m happy to share that this deep venture culminated in to a book, “A Practical Guide to Molecular Docking”, published with Springer Nature, will soon be available (Tentatively July 2026).

This book is not just about how to perform docking, but about how to think about docking critically.
It covers:
- Theoretical foundations of molecular interactions
- Practical workflows using widely used docking tools
- Common pitfalls that lead to misleading conclusions
- Integration with molecular dynamics and free-energy methods
- Reproducibility and best practices in docking studies

The goal was simple, to bridge the gap between textbook docking examples and real-world drug discovery challenges. This insight, which became increasingly clear during drafting and the literature review that enabled me complete the manuscript in to a coherent piece.

I’d be very interested in perspectives from others working in this area:

The book will be soon available for pre-order here: https://lnkd.in/dw_htfFt

PS: Link for amazon will be shared later.

The ergodic hypothesis occupies a foundational role in statistical mechanics by providing the conceptual bridge between microscopic dynamics and macroscopic observables. In principle, thermodynamic properties—such as energy, pressure, or entropy—are defined as ensemble averages over an enormous number of microstates. However, in practice, both experiments and molecular dynamics simulations access only a single system evolving over time. The ergodic hypothesis justifies this substitution by asserting that, over sufficiently long times, the time average of an observable along a trajectory becomes equivalent to its ensemble average.

This equivalence is what allows molecular dynamics simulations to compute meaningful physical properties from a single trajectory, rather than requiring explicit sampling of all possible configurations. Without this assumption, the connection between simulation outputs and thermodynamic quantities would be fundamentally unclear.

Importantly, ergodicity is not guaranteed for all systems or timescales. Many realistic systems—especially biomolecules with rugged energy landscapes—may exhibit slow dynamics, metastability, or incomplete phase space exploration, leading to deviations from ideal ergodic behavior. Therefore, while the ergodic hypothesis provides the theoretical justification for simulation-based averaging, its practical validity must always be critically assessed in terms of sampling efficiency and convergence.

I hope you liked and enjoyed this video.

Please subscribe to my YouTube channel: @EMs_MDLabs

In this video, we understand the crux of computer simulations, the potential energy surface. A potential energy surface (PES) in biomolecular simulations is the high-dimensional hypersurface that maps molecular configurations to their corresponding potential energies. Each point on the PES represents a specific arrangement of atomic coordinates R, and the energy is typically computed using a molecular mechanics force field (e.g., bonded terms for bonds, angles, torsions, and non-bonded terms for van der Waals and electrostatics).
For a biomolecule with N atoms, the PES exists in a 3N-dimensional space (reduced by constraints such as fixed bond lengths). Local minima correspond to stable conformations (e.g., folded states, metastable intermediates), while saddle points represent transition states governing conformational changes. The topology of the PES—its wells, barriers, and pathways—determines thermodynamic stability and kinetic accessibility.
In practice, biomolecular simulations (e.g., molecular dynamics) do not construct the full PES explicitly but sample it dynamically via trajectories. Efficient exploration of relevant regions of the PES is critical, as the surface is rugged with numerous local minima, particularly for large systems like proteins.

I hope you enjoyed this video.
Please subscribe to my YouTube Channel: @EMs_MDLabs.

In this second video from the molecular dynamics 101 series, I explain the importance of force fields, which are central to MD simulations. The source of accuracy and errors both stems from the force field parameters. Developers meticulously fine-tune the parameters over time, incorporating new results and discarding obsolete or erroneous parameters. 

I hope you learn and enjoy this new video.

Don’t subscribe to my YouTube Channel: @EMs_MDLabs.

Subscribe now

I have created a YouTube Channel, where I will simplify computer simulations. This topic has been close to my heart, and I have working on this for 10 years now. The idea to simplify complex topics that haunts beginners who wish to learn MD simulations. I am no PRO in this but I wish to share my learnings and experiences. As it is famously said, “Teaching is learning twice“, I intend to hone my knowlegde and skills along the way.
The intial set of videos on YouTube are AI-generated using NoteBookLM using classical MD and thermodynamics textbooks (Some are rare to find these days). These are supplemented by state-of-the-art good practices papers published in last 2-5 years.
I hope this series serves to educate those aspiring to learn biomolecular simulations.
Happy Learning.

Please send in your request for contents and topics on eafmartis@gmail.com.
Follow me on Twitter @EMs_MDLabs.

Click here to subscribe to my YouTube Channel