Since the development of the first supercomputers in the sixties, high performance computing has become a pillar of modern science. In fields as diverse as particle physics, cosmology, meteorology, materials science, geology, and genetics, computational simulations are employed to explore the consequences of complex mathematical models, predict the outcome of relevant experiments, and ultimately bridge the gap between theory and observation. Continuous advances in technology have brought the processing capacity of individual supercomputers to levels comparable to, if not exceeding, those of the human brain (1). This exascale computing revolution is catalyzing a renaissance in artificial intelligence: computational tasks such as machine learning and data mining have reached maturity, enabling the discovery of fundamental physical laws at the LHC (2) and novel patterns in medical data (3).
The enormous computational advances of recent years promise to fulfill the molecular biophysicist’s dream: to interpret complex physiological phenomena in light of a quantitative microscopic theory. This bottom-up approach to the life sciences—from atoms to tissue—to address the complexity of living systems through the synergy between chemistry and statistical physics, is now at our fingertips. Recent advances in experimental techniques, including cryo electron microscopy (4) and single molecule imaging (5), together with improved computer hardware (6) and molecular dynamics algorithms (7), continue to shrink the gap between experiment and theory.
Among the key molecular phenomena underlying human physiology, ion translocation constitutes a critical factor for the emergence of life (8). Importantly, ion transport is at the basis of cellular excitability, and was one of the first biological problems addressed by mathematical modeling. The transformative insight of Hodgkin and Huxley was the first milestone in a journey culminating with endeavors like the Human Brain Project (9) and Brain (10). Yet even these ambitious projects may not represent the apex of the field: in the pharmaceutical industry, GSK is currently investing in bioelectronics to provide proof of concept for electroceuticals: implantable nanodevices restoring function to tissues and organs by means of action potentials transmitted to peripheral nerves (11). Such innovative technologies are accompanied by novel challenges, from designing biocompatible electronics materials able to interact with cellular components, to predicting the net electrical signals ensuing in neurons. To meet these demands, we will have to address the intrinsically multiscale problem of ion transport in accurate and efficient ways.
Unfortunately, we are still far from a seamless theory of everything for ion transport and cellular excitability. Attempts at bridging the time- and length-scale divides relevant to these problems have been based mostly on three core ideas: i) developing coarse-grain potentials to describe the dynamics of a subset of degrees of freedom, thereby increasing the time- and length-scales that can be explored; ii) non-canonical probability distribution sampling to speed up exploration and reconstruct the dynamics of relevant experimental observables; and iii) modeling a system's dynamics using large discrete-state kinetic networks to make inferences over long time scales.
No single approach provides a satisfactory quantitative connection between atom-based calculations and cell- or tissue-level responses, which are usually characterized in terms of phenomenological models. Whereas synergy among these components is highly desirable, progress in this direction has been hindered thus far by a lack of common background and language.
In this workshop, we aim to unite a diverse community of scientists focused on computational simulations as well as functional characterization of ion transport across membranes, to discuss challenges and current limitations and to identify possible solutions across methodological boundaries. Our proposal is motivated by the observation that, whereas molecular simulations have begun to provide satisfactory answers concerning equilibrium properties of biomolecules, modeling kinetics remains a major challenge. Specifically, several questions, including the mapping between coarse-grain and all-atom dynamics, the modeling of kinetics from a potential of mean force, or the amount of conformational space exploration needed to initialize Markov state models, have not yet received definite answers. The workshop will focus on the major open questions in the field and on the approaches used to address them from various angles..
Our overarching goal for this workshop is to develop a common vision of computational needs for the future of biological ion channel research. At present, several countries are developing strategic plans for leadership in computing machines (12). Due to the current explosion in hardware heterogeneity, computational power does not necessarily imply efficiency in solving a specific numerical problem. Careful planning is required to design machines that are not only powerful, but also suitable for solving relevant scientific problems. Accordingly, scientists from all fields will have to play a crucial role in shaping the machines of tomorrow. Biophysics faces a peculiar challenge in that heterogeneous approaches (kinetic network models, coarse-grain systems, all-atom simulations) produce conflicting hardware requirements (e.g. capacity vs. parallel scalability vs. processor and I/O performance). By facilitating communication, this workshop will help researchers in the ion channel community to define common goals and thus contribute to the ongoing planning of next generation supercomputers.
We envision four themes as major topics of discussion:
1. Identify roadblocks limiting present success of atomistic modeling. Why do molecular simulations sometimes fail to give correct answers? How can these challenges be effectively communicated beyond the simulation community?
2. Share success stories and analyses thereof: What are key elements of successful modeling?
3. Address the technical language divide by clarifying common language for communities specializing in computational simulations and functional characterization.
4. Summarize and disseminate best practices and recommendations for future studies.
The following sponsors are acknowledged for their support of the workshop:
BioExcel is a Centre of Excellence providing support to academic and industrial researchers in the use of high-performance and high-throughput computing in biomolecular research. BioExcel-sponsored Workshop Leads will guide hands-on sessions during this CECAM program on key applications of computational methods to modeling ion transport.
The Journal of General Physiology (JGP) aims to publish mechanistic and quantitative molecular and cellular physiology of the highest quality, to provide a best-in-class author experience, and to nurture future generations of independent researchers. JGP will contribute an iPad Mini to the winner of the Poster Competition at this CECAM program.