Phase Space Invaders (ψ)
With the convergence of data, computing power, and new methods, computational biology is at its most exciting moment. At PSI, Miłosz Wieczór asks leading researchers in the field to discover where we’re headed for, and which exciting pathways will take us there. Whether you’re just thinking of starting your research career or have been computing stuff for decades, come and join the conversation!
Featuring links to all episodes and exclusive summaries containing some of the key messages, highlights and learning points. All content created by Miłosz Wieczór.
Episode 1 features an interview with Pilar Cossio, a project leader at the Flatiron Institute working on computational methods to integrate experimental data, such as cryo-electron microscopy or single-molecule spectroscopy.
Pilar discusses her path from studying physics in Colombia to doing her PhD on computational protein folding, followed by postdocs developing methods to extract information from experiments like cryo-EM and force spectroscopy. She highlights the field’s excitement about the increasing interplay between simulations and experiments, with theorists now helping guide experiments rather than just interpreting results after the fact.
One key area Pilar is working on is in situ cryo-EM tomography, which involves flash-freezing entire cells rather than purified proteins. This allows viewing biomolecules in their native cellular environments, but presents challenges like identifying the molecule of interest among the crowded, noisy cellular landscape. While techniques are rapidly advancing, even simply locating a protein of interest within the frozen cell remains a difficulty, so more capable theoretical methods need to be developed.
The conversation also touches on potential pitfalls of the increasing prevalence of social media in science. Pilar worries about science becoming more superficial and journals favoring flashy results over deep explorations. She also ponders whether the huge number of publications could be counterproductive information overload, and we briefly discuss possible solutions to this.
Finally, Pilar points out that code development and maintenance is undervalued compared to publishing new methods. Many important community codes rely on dedicated developer teams, but securing long-term funding for this behind-the-scenes work is challenging, leading to persistent bugs or software losing functionality over time. Overall, the technology enabling new scientific advances also creates new challenges around data management, communication practices, and incentive structures within the academic culture.
Episode 2 features an interview with Max Bonomi, a group leader at Institut Pasteur in Paris who has been deeply involved with foundational tools and methods for molecular simulations and integrative biophysics.
Max discusses how his work on developing the widely-used Plumed plugin for molecular dynamics simulations stemmed from efforts during his PhD to unify implementations of enhanced sampling methods like metadynamics across different codes. He expresses excitement about the potential of AI models to not just predict protein structures, but to characterize full conformational landscapes including populations, interconversions, and thermodynamics/kinetics – going beyond what current molecular dynamics can efficiently provide.
Max sees opportunities for AI to learn from and augment physics-based simulations and experimental data. He is also enthused by experimental advances like in situ cryo-EM tomography revealing biomolecules in cellular environments at improving resolutions.
The conversation turns to undervalued aspects of scientific work. Max laments that efforts like software development, documentation, tutorials, and other services to the community are not as well-recognized as publications and talks, especially for early-career researchers. He advocates for finding better ways to value and incentivize these crucial contributions beyond traditional metrics.
Max and Miłosz also discuss the challenges of appropriately crediting collaborative work, peer review, and other scientific activities as the community grows larger and more open. Overall, Max highlights the importance of holistically supporting all aspects of scientific training and work beyond just publishable research outputs.
Episode 3 features an interview with Wojtek Kopeć, a lecturer in computational pharmaceutical chemistry who recently moved from his post-doctoral position at the Max Planck Institute to start a new group at Queen Mary University of London.
Wojtek discusses the challenges of starting a new research group, such as suddenly being responsible for undergraduate teaching, navigating unfamiliar university bureaucracies and weirdly sounding acronyms, as well as finding little time for his own research amidst myriad administrative tasks. He reflects that the transition from a pure research institute to a teaching-focused university was a major adjustment.
The conversation then turns to the evolving relationship between academia and industry in science. Wojtek points out how some longstanding computational challenges like protein folding are increasingly being solved by deep-pocketed companies rather than academia. He ponders whether academia’s future role may shift more toward training and collaboration with industry.
Wojtek and Miłosz comment on the lack of recognition for crucial scientific activities like software development, documentation, and peer review compared to publications. They advocate for better valuing these often underappreciated contributions, especially for early-career researchers.
Wojtek highlights the excitement about the growing interplay between computational and experimental approaches, allowing simulations to guide experiments and provide relatable observables. However, he raises concerns about occasional statistical lapses in computational studies getting published without rigorous error analyses.
Overall, Wojtek highlights the increasingly social and interdisciplinary nature of modern science, diverging from the myth of the lone genius scientist. He calls for finding better ways to promote productive collaborations and give due credit for all vital scientific efforts beyond just research outputs.
Episode 4 features a conversation with Modesto Orozco, a renowned researcher known for his contributions to the multiscale study of nucleic acids. Modesto shares his journey from starting as a quantum chemist to expanding his research to encompass various levels of physics modeling and experimental findings in computational biology and biophysics.
The discussion revolves around the exciting period we are currently in, where computational research can drive experiments and provide testable hypotheses, rather than merely reproducing known facts. Modesto emphasizes the importance of open data and transparency in science, even if it means making mistakes public. He believes that the upcoming revolution in biology relies on integrating different types and massive amounts of experimental and computational data.
Modesto reflects on how the field has evolved from simply reproducing known phenomena to providing rationales for observed behavior, and now towards predictive power. He cites examples like the prediction of RNA folding and stability, which can fuel the design of more powerful vaccines. Modesto envisions a future where theoretical calculations will become a central resource for further developments, akin to structural genomics databases.
The conversation then shifts to the need for integrating data and information from various sources. Modesto underscores the importance of sharing simulation data and trusting the results, even if they may be flawed, because this way the community can identify and address the flaws. He believes that making data available in a good format can benefit diverse areas, from drug design to biophysics studies and cell biology, fostering collaborations and fueling new discoveries.
Overall, the episode highlights the transformative potential of computational approaches in biology and the importance of open data sharing and interdisciplinary collaboration to drive scientific progress.
Episode 5 features a conversation with Paul Robustelli, an assistant professor of chemistry at Dartmouth College who specializes in modeling intrinsically disordered proteins. Paul shares his journey from studying NMR data as an undergraduate to exploring the world of disordered proteins during his PhD and postdoc.
The discussion starts with Paul’s experience working at D.E. Shaw Research, a private institution known for its bold goals and unconventional funding model. He highlights how this exposure to a different funding structure informed his understanding of resource allocation and influenced his subsequent grant writing process.
Paul also reflects on the evolving landscape of science funding, with the rise of computational startups, private philanthropic funding, and hybrid models like the Open Force Field Consortium. He expresses optimism about these diverse funding models, believing they can provide more options and alleviate the bottleneck of traditional academic career paths.
Shifting gears, the conversation explores the importance of embracing the multidimensional nature of scientists. Paul emphasizes the value of fostering diverse interests and hobbies, drawing parallels between the scientific community and artistic pursuits. He encourages viewing scientists as well-paid artists, celebrating their creative spirits and the richness of perspectives they bring to their work.
Overall, the episode highlights the blurring boundaries between academia and industry, the evolving funding models in science, and the importance of recognizing scientists as multifaceted individuals with diverse passions beyond their research.
Episode 6 features an insightful conversation with Giulia Palermo, a prominent group leader at the University of California Riverside, renowned for her pivotal contributions to understanding the molecular mechanisms of the CRISPR-Cas9 system.
The discussion begins with Giulia sharing her journey of becoming involved in the CRISPR-Cas9 research, recounting how she was intrigued by the system’s potential to revolutionize biotechnology and her determination to unravel its intricate workings through computational approaches, despite initial skepticism from reviewers.
Giulia highlights the multi-scale nature of the Cas9 system, which requires integrating various levels of description, from quantum chemistry to large-scale conformational changes and binding events. She emphasizes the need for computational methods to adapt to the increasing complexity revealed by cutting-edge experimental techniques like cryo-electron microscopy.
The conversation then turns to the broader revolution in RNA science, where new discoveries are constantly shifting paradigms in cell biology. Giulia shares her excitement about the diverse mechanisms and therapeutic strategies emerging from RNA research, which challenge traditional drug discovery approaches focused on receptor binding.
Giulia also underscores the importance of embracing creativity in scientific pursuits, arguing that transformative breakthroughs often arise when researchers venture outside their comfort zones and challenge conventional thinking. She advocates for maintaining “pet projects” that may take years to yield publishable results but have the potential to drive paradigm shifts.
Furthermore, Giulia highlights the increasing relevance of microbiology in uncovering novel biological functions, challenging the assumption that bacterial systems are less complex than human cells. She cites examples from the CRISPR-Cas family, illustrating the diverse applications beyond gene editing, such as nucleic acid detection and imaging.
Overall, the episode offers insights into the challenges and opportunities in computational biology, emphasizing the need for integrative approaches, creative thinking, and a willingness to explore unconventional avenues to drive transformative scientific discoveries.
Episode 7 features Aleksei Aksimentiev, a group leader at the University of Illinois Urbana-Champaign whose research focuses on computational studies of biological and bio-inspired nanosystems. Aleksei shares his unconventional path from studying particle physics in Ukraine to a PhD in polymer physics in Poland, before serendipitously discovering and transitioning into biophysics and biomolecular simulations.
He recounts the funny story of how a single Google search eventually led him to join Klaus Schulten’s lab, catalyzing his foray into modeling biological systems like ATP synthase. A chance encounter with nanotechnologist Greg Timp sparked Aleksei’s interest in nanopore sequencing, where he did pioneering simulations of DNA translocation through solid-state nanopores. This highly multidisciplinary research interfacing biology with nanotechnology became a specialty of his lab.
Aleksei discusses milestones like the realization of nanopore DNA sequencing, protein sequencing as a next frontier, and his group’s modeling of DNA nanostructures. He emphasizes the importance of being open to new fields and leveraging an environment conducive to collaboration across disciplines.
The conversation shifts to the indispensable role of mentorship in academia. Aleksei credits his PhD advisor Robert Hołyst for inspiring his scientific curiosity, and post-doctoral advisor Klaus Schulten for insights into the granular realities of publishing and funding. He stresses mentorship as a two-way endeavor built on the mentor’s ability to ignite passion, and the mentee’s hunger to learn the unwritten nuances of scientific careers.
Episode 8 features Rossen Apostolov, the director of the BioExcel Center of Excellence affiliated with KTH Royal Institute of Technology in Stockholm. Rossen coordinates a multi-faceted European effort to bring computational biophysics communities closer together through code development support, training, and promoting interoperability between specialized software.
Rossen discusses the importance of scientific collaboration amidst increasing complexity of computational resources and experimental data. He highlights how centers like BioExcel avoid redundant efforts by facilitating knowledge sharing and following best practices for software sustainability.
The conversation covers BioExcel’s work on core applications like GROMACS and HADDOCK, developing integrative modeling workflows, and producing high-quality training materials and webinars. Rossen emphasizes balancing powerful software with user-friendliness to increase adoption.
He shares insights on managing a large consortium, securing long-term funding by delivering indispensable tools/services, and the benefits of such initiatives extending beyond the core partners. BioExcel’s ambassador program aims to nucleate local communities across Europe.
Rossen sees his role as ensuring a clear vision while onboarding new members excited to collaborate. The greatest rewards are emails from students aided by BioExcel’s training in advancing their careers, creating a “family” of computational biophysicists with shared foundations.
Episode 9 features Michele Vendruscolo, a professor of biophysics and group leader at the Centre for Misfolding Diseases at the University of Cambridge.
In the conversation, Michele discusses his group’s work using computational methods to study protein misfolding and aggregation, which are involved in neurodegenerative diseases like Alzheimer’s and Parkinson’s. He starts with some background on how he originally got interested in protein folding as a physics student, before shifting his focus to protein misfolding and its role in disease during his postdoctoral research, initially working as a computational biophysicist in a fully experimental lab. Michele explains that while the amyloid hypothesis has been debated for decades, the recent approval of antibody drugs targeting amyloid plaques provides validation that protein aggregation is a causative agent worth targeting therapeutically.
To temper the enthusiasm, though, he notes these approved antibodies still have significant side effects and high cost, limiting their use so far. His group is now exploring using AI methods to design small molecule drugs, so-called molecular chaperones, that can stabilize the native, functional state of amyloid proteins and prevent their misfolding and aggregation. Michele is optimistic that some of these computationally designed drugs could start to transform the clinical practice within 5-10 years.
As the conversation shifts to the role of public engagement for scientists, Michele argues there is a duty for researchers to communicate their work to the public, especially in areas with potential impacts on human health and global challenges. He encourages young scientists to seek out good mentors and to direct their efforts towards fields where their contributions are truly needed by society. Reflecting on his own career path and lab trainees, Michele highlights the crucial importance of strong scientific mentorship. He advises young people feeling the urge to understand nature through science to channel that curiosity towards working on humanity’s most pressing challenges.
Episode 10 features Ariane Nunes Alves, a group leader at Technische Universität Berlin known for her work on methods to estimate the kinetics of ligand binding and unbinding. While drug design often focuses on binding affinity, Ariane emphasizes the importance of considering binding kinetics as well: cells exist in non-equilibrium states, so kinetic rate constants better describe cellular processes than equilibrium measurements like dissociation constants. Though computationally more demanding, understanding binding pathways provides insight into time-dependent drug behavior.
Ariane most often works on reproducing relative kinetic data in her simulations, giving confidence that key biological details are captured, if not the absolute rate values. However, she points out that good experimental probes are lacking to verify the actual simulated binding pathways, especially for single binding pockets. The discussion shifts to simulating biologically realistic cellular environments, which are highly crowded with proteins and other macromolecules. As shown recently, molecular crowding can e.g. significantly alter the binding pathway between a ligand and protein target.
On the social aspects of being a principal investigator, Ariane argues that a major part of the job involves interpersonal interactions through teaching, mentoring students, and collaborations. Yet hiring committees often overlook evaluating these critical “softer” skills. She brings up stories of students treated poorly by supervisors lacking emotional intelligence, and makes a case that such issues should be treated seriously.
The conversation concludes pondering the lack of a central online community platform for computational biophysicists following the disruption of Twitter. While different platforms like Mastodon and LinkedIn have emerged, none has recreated the informal gathering spot Twitter provided before Elon Musk’s takeover. We speculate that maintaining connections may require a new virtual “meeting place” for exchanging ideas.
Episode 11 features Justin Lemkul, an assistant professor at Virginia Tech working on polarizable force fields for nucleic acids and proteins. Justin is well known for his extensive online efforts creating tutorials, answering questions on forums, and being a go-to problem solver for computational biology.
The discussion begins with Justin recounting how his quest to help others online stemmed from his own struggles as a student trying to learn molecular simulation software with limited resources. Over 15 years, he has answered thousands of questions across forums and mailing lists, developing insights into effective scientific communication tailored to different experience levels and backgrounds.
While tutorials increase accessibility, Justin cautions that simulations require deep specialist knowledge to avoid treating them as black boxes. He advocates communicating to students and PIs that high-quality simulations are real computational science requiring critical thinking, not just black-box tools to obtain answers easily.
Turning to his research, Justin makes a case for using polarizable force fields to better model the highly charged and dynamic nature of nucleic acids, especially non-canonical structures like G-quadruplexes. In such structures, explicit polarization captures effects like electronic induction between bases and ion coordination that additive models struggle with. His work has expanded from DNA to exploring polarizability’s impact on catalytic RNA ribozymes and folding.
As experimental RNA biology undergoes a renaissance, Justin sees an opportunity for predictive simulations to drive hypotheses and enlighten mechanisms in this underexplored frontier. However, he acknowledges the paradox that simulations may precede validation data for many systems. Nevertheless, he encourages using simulations’ predictive power to motivate future experiments and iteratively refine models.
The conversation closes reflecting on how to effectively convey the rapid discoveries and remaining uncertainties around RNA’s diverse roles to the next generation of students through redesigned curricula and courses that balance decades-old simplifications with the latest research findings.
Episode 12 features an interview with Vlad Cojocaru, a computational biologist working on transcription factors and their role in cellular reprogramming. The conversation revolves around the current challenges and future prospects of modeling transcription control and genome organization in human cells. Vlad discusses the increasing scale and complexity of biological systems being modeled, from simulating transcription factors binding to DNA to their interactions with nucleosomes and the epigenetic landscape. He envisions future models that can tackle the rearrangement of multiple nucleosomes by transcription factors, capturing the intricate dynamics of gene regulation.
The discussion touches upon the need for multiscale modeling approaches, combining atomistic simulations with coarse-grained methods to capture large-scale motions and systems. Vlad highlights the potential of AI and machine learning to predict structural dynamics of nucleosomes and their arrangements, given the availability of sufficient data.
The conversation then shifts to Vlad’s decision to return to his home country, Romania, after years of research abroad. He weighs the pros and cons, including the potential benefits of a permanent position, the challenges of limited funding and resources, and the opportunity to contribute to the scientific landscape in Romania. Vlad emphasizes the importance of fostering international collaborations and attracting talent from around the world, rather than solely focusing on retaining local researchers.
Finally, Miłosz and Vlad discuss the broader implications of “brain drain” narratives and the need to reshape the discourse towards attracting researchers back to their home countries. They argue that a critical mass of returning scientists can drive positive change, promote postdoctoral training, and cultivate a culture of international collaboration, ultimately benefiting scientific progress and societal development.
Episode 13 features Daniel Zuckerman, a professor of biomedical engineering at Oregon Health and Science University School of Medicine. Daniel is known for his contributions to statistical molecular biophysics and as an avid popularizer of a physical view of biological processes.
Daniel discusses his motivation for sharing ideas through writing, tracing it back to his background in humanities before switching to physics. He explains how this unique perspective helped him develop accessible materials for students from diverse backgrounds learning statistical biophysics.
The conversation explores the landscape of statistical biophysics book writing, with Daniel emphasizing the importance of starting from simple instances to reach general principles. He highlights the value of a physics-informed narrative in unlocking the understanding of biology, using examples like the connection between free energy expenditure and proofreading mechanisms in cells.
Daniel shares his views on the current state of computational molecular biophysics, advocating for more focus on sampling and cautioning against overreliance on increasingly large simulations without proper statistical analysis. He discusses the trajectory-based methods in biophysics, particularly weighted ensemble simulations, explaining their strengths in studying non-equilibrium processes and their pedagogical value.
The episode also touches on the potential integration of machine learning techniques with trajectory-based methods to enhance sampling processes and optimize reaction coordinates. Daniel mentions ongoing work in this area, including efforts to minimize statistical uncertainty in rate constant estimations.
What stands out throughout the conversation is Daniel’s passion for clear communication and rigorous scientific approaches in the field of statistical biophysics, making this episode an insightful look into the intersection of physics, biology and teaching.
Episode 14 features Syma Khalid, a professor of computational microbiology at the University of Oxford. Syma is renowned for her pioneering work on understanding the bacterial cell envelope through computational methods.
The conversation begins with Syma explaining the complexity and importance of the bacterial outer membrane. She describes it as a highly impermeable “coat of armor” that protects bacteria while selectively allowing essential nutrients to enter. Syma discusses recent discoveries about the spatial organization of proteins and lipids in the membrane, which had previously been considered random.
Syma shares insights on potential new approaches to antibiotic design, including learning from how bacteria and bacteriophages naturally attack each other. She highlights the “Trojan horse” approach as a promising strategy, where harmful elements are attached to molecules that bacteria naturally uptake.
The discussion then shifts to Syma’s philosophy for running her research group. She emphasizes the importance of creating a relaxed and flexible work environment, encouraging social activities, and supporting her team members’ personal growth and happiness. Syma stresses that there’s no hierarchy between academia and industry, and that contentment with their chosen life path should be the ultimate goal for her students and postdocs.
Syma also touches on the value of diversity in research groups, noting how it fosters empathy and emotional intelligence. She shares a lesson learned from her mentor about prioritizing kindness in the academic environment. The episode concludes with reflections on the importance of focusing on overall well-being in addition to scientific productivity.
Episode 15 features Paulo Souza, group leader at the Laboratory of Biology and Modeling of the Cell at the École Normale Supérieure in Lyon. Paulo is the main developer of Martini, a popular coarse-grained force field used in computational biology.
Paulo discusses the challenges of developing a universal force field for biology that incorporates empirical knowledge from various domains. He explains Martini’s “building block” approach, which uses Lego-like pieces to construct models of biomolecules. The development process starts with basic components and progresses to more complex systems, allowing for comprehensive coverage of chemical interactions.
The conversation touches on the limitations of coarse-grained modeling, including the need to compensate for reduced degrees of freedom and the challenges of representing certain molecular behaviors. Paulo emphasizes the importance of community feedback in refining the model and addressing shortcomings, such as the “stickiness” issue in protein-protein interactions that was improved in Martini 3.
Paulo also discusses the ongoing efforts to reduce reliance on elastic network models and allow proteins more freedom in their behavior. He mentions the potential impact of AI and machine learning on force field development, particularly in areas like automatic parameterization and improving bonded potentials.
The episode concludes with a discussion on the future directions of Martini development, balancing the study of small systems with modeling larger, more complex biological structures. Paulo highlights the importance of both approaches, noting that while complex systems offer more realistic biological applications, simpler systems are crucial for model validation.
Episode 16 features Janusz Bujnicki, head of the Laboratory of Bioinformatics and Protein Engineering at the International Institute of Molecular and Cell Biology in Warsaw. Janusz discusses his journey from experimental biology to bioinformatics at the turn of the century, when the first bacterial genomes were being sequenced. He explains how his lab now integrates bioinformatics, structural biology, and experimental approaches to study complex biological questions.
Janusz explains the challenges of RNA structure prediction, emphasizing the need to combine physics-based and data-driven approaches. He highlights his group’s work on the Modomics database of RNA modifications and their efforts to incorporate these modifications into structural predictions. Janusz also touches on his participation in RNA structure prediction competitions like CASP and RNA Puzzles.
The conversation then shifts to Janusz’s experience as a scientific advisor for policymaking at the European level. He explains the distinction between science policy, science diplomacy, and scientific advice for policymakers. Janusz emphasizes the importance of interdisciplinary collaboration when providing scientific input on complex societal issues. He illustrates this point with an example of advising on gene editing versus genetically modified organisms in agriculture, highlighting the challenges of translating scientific knowledge into policy decisions.
Throughout the episode, Janusz underscores the importance of integrating different approaches in scientific research and the role of scientists in informing policy decisions while maintaining a sense of humility about the broader context in which these decisions are made.
Episode 17 features Caroline Lynn Kamerlin, a professor of molecular design at Georgia Tech and professor of structural biology at Uppsala University who primarily studies catalytic mechanisms of enzymes. The conversation covers a wide range of topics, reflecting Caroline’s broad interests and interdisciplinary background.
Caroline discusses the importance of asking bold scientific questions and engaging in truly interdisciplinary research. She attributes her ability to work across disciplines to her background in natural philosophy. The discussion delves into emerging concepts around enzyme promiscuity, exploring how this view has become widely accepted in the field and its implications for research and human health.
A significant portion of the conversation focuses on the growing problem of bacterial resistance in public health. Caroline expresses concern about the lack of public awareness regarding antimicrobial resistance and emphasizes the need for more research and policy attention in this area.
The episode also touches on science education and communication. Caroline highlights the importance of teaching critical thinking skills to combat misinformation and conspiracy theories. She stresses the responsibility of scientists to engage in public communication and the need for a more proactive approach to addressing global challenges like climate change and pandemic preparedness.
Throughout the discussion, Caroline shares insights from her experience writing commentary articles and opinion pieces on various scientific and societal issues. She emphasizes the value of interdisciplinary collaboration and the need for spaces where researchers from diverse fields can interact and exchange ideas.
Episode 18 features Erik Lindahl, professor of biophysics at the University of Stockholm and KTH Royal Institute of Technology. Erik, known for coordinating the GROMACS project, discusses his research on pentameric ligand-gated ion channels and shares insights on the future of molecular biophysics.
Erik emphasizes the importance of thinking and having smart ideas in science, rather than relying solely on computational power. He believes that the human element – the ability to come up with novel ideas and approaches – will remain crucial in biophysics for decades to come.
The conversation delves into Erik’s work on ligand-gated ion channels, highlighting the complexity of these structures and the challenges in understanding their function. Erik explains that while we grasp the basic principles of these channels, there are still many unknowns, particularly regarding the process of desensitization and recycling back to the resting state.
Erik discusses the value of combining computational methods with experimental techniques, particularly electrophysiology. He shares an anecdote about how simulations can sometimes lead to unexpected results, emphasizing the importance of verifying computational predictions with experimental data.
The podcast also touches on the topic of data quality and accessibility in scientific research. Erik advocates for better quality assessment indicators for computational data and highlights the importance of making research data publicly available.
Finally, Erik talks about his series of online lectures on concepts in molecular biophysics, which he created during the COVID-19 pandemic. These lectures, available on YouTube, have reached a wide audience and serve as a valuable resource for students and researchers in the field.
Throughout the conversation, Erik emphasizes the importance of simplicity in scientific thinking and encourages young researchers to share their knowledge and contribute to the field, regardless of their career stage.
Episode 19 features a conversation with Alex MacKerell, a professor of Pharmaceutical Sciences and director of the Computer-aided Drug Design Center at the University of Maryland School of Pharmacy in Baltimore. Alex is renowned for his contributions to the CHARMM family of atomistic force fields, including innovations like dihedral correlation CMAP terms and the polarizable Drude particle model.
Throughout the discussion, Alex advocates for a pragmatic approach to force field development, emphasizing the concept of “big molecules, small physics.” He highlights the importance of combining machine learning with human intelligence in model development, cautioning against over-reliance on automation which can lead to unphysical and non-transferable parameters. Alex discusses ongoing efforts to expand the coverage of force fields into diverse chemical spaces, including drug molecules and complex biological systems. He emphasizes the need for careful parameterization, particularly for ion-biomolecule interactions in polarizable force fields.
The conversation also touches on the challenges of obtaining high-quality experimental data for force field development. Alex points out the reliance on decades-old physical chemistry measurements and the need for new experiments in areas like hydration free energies and osmotic pressures.
Finally, Alex stresses the importance of standardized protocols and tools like CHARMM-GUI for ensuring reproducibility in molecular simulations. He discusses the challenges of implementing new force field developments across different software packages and the potential benefits of creating repositories for consistently generated trajectories to improve reproducibility and facilitate machine learning applications.
