Property:Post-meeting notes
From Complex Time
This is a property of type Text.
A
<u>'''Talks'''</u>
'''Matteo Osella'''. Interesting idea of connecting laws of physiology (Hwa) with aging/senescence. Not trivial how to do that for single cells.
'''Lin Chao'''. Aging and asymmetry in E. coli. Advantage of asymmetry is portfolio diversification. Somewhat optimal level of asymmetry emerges.
'''Uli Steiner''' . Fitness as combination of fecundity and mortality. Death in the mother machine (surprisingly high): mother (early daugther) has an increased mortality rate with age, while her latest daughter has an approximately constant mortality rate. Idea: late daughter inherits the damage, while the mother was starting with minimal damage. No correlation between mother and late daughter lifespan.
'''Sri Iyer-Biswas.''' Cool data on C crescentus and collapses. Interesting observation of memory of past conditions lasting for long time.
'''Owen Jones.''' Senescence across the tree of life. Measure shape and pace (timescale)
'''Sabrina Spencer.'''
'''Bree Aldrige'''
'''Chris Kempes'''
'''Martin Picard'''
'''Geoffrey West'''
'''Ideas'''
What is aging? Requires asymmetry in division and the ability to label individual with a "time stamp". In E. coli age of the pole, in mycobacteria cell wall. Senescence is the loss of function associated to aging. The question then is what is function. We have a bias for growth rate.
It is very unclear to my whether asymmetry is adaptive or not. It is also unclear how to prove it.
The other axis is memory. Memory (information) about the environment. Unclear how that is related with aging. +
C
A number of critical questions were raised about the best levels at which to establish causality when it comes to understanding both natural and disease-related aging. Namely what are the best observables to consider? Should these be single measurements or network based measurements. Could the best indicators involve comparisons across genetic and cognitive networks applying similar methods, or as is more typical time-dependent changes in a given network at one level of analysis. A recurring question was the relationship between energy and information and how their reciprocal dependencies change over the course of time and the course of disease.
Some very general issues that arose in conversation that require further exploration include:
#Approaching disease from a first-principles theoretical perspective - as is common in ecology - thus establishing principled data collection objectives (this would require a rigorous operational definition of the disease state in formal terms)
#The value and limitation of the current inductive, big data approach, that focuses on time-dependent associations
#The meaning of cognitive reserve, exercise or error correction, and the limits to these
#How adaptive phenomena that are ongoing mitigate the disease state or at some point perhaps accelerate it.
#How we might better explore causality in large systems with extensive non-linear feedback mechanisms. +
H
A theme that I see across the talks and in the discussion is the issue of complexity and how integrity of complex systems is lost with age and how it might be retained to impinge on health and resilience.
The idea of the adaptive landscape is very useful as is the idea of tipping point - i particularly like the idea of aging as a series of transitions where the path taken dictates the possibilities open for the future
Hochberg: Concepts that caught my attention, as a function of age is loss of resilience equally felt through the lifespan ie young v old? Also Diverse/idiosynchratic networks – how should models be informed. The idea of hierarchies of regulatory or adaptive nodes is interesting but I wonder do we know that there are grades of nodes in the first place, if there are how do we find them?
Crespi : Biological Risk Matrix… coupling versus complexity. I had difficulty with this idea because the systems were assigned importance but it wasn't clear to me what the basis for those assignations was. I have viewed the organ systems as different but inseparable pieces of the organism as a whole. I do like the idea of viewing the aging of specific processes in terms of trade-offs - I wonder about the inbuilt redundancy of systems and think we could consider the possibility that age-induced adaptions might just as well be beneficial - ie tailored to the prevailing internal environment or the current disposition of regulatory nodes.
Valenzano: The fact that ecology predicts genome size was super interesting and that the expansion is explained by transposons! I also loved the idea that the long-lived species had more emphasis on positive/purifying selection and the short-lived species had more evidence of the relaxed selection. These are an amazingly useful species for the interactions of genetics, environment - the exposome!
Di Gregorio: Among cancers age dependence in risk is shared despite differences in etiology and mechanisms of tumorigenesis and differences in the stem cell pools that these cancers arise from. All map to a common cancer curve – incidence as a function of age all lie right on top of each other. This nicely captures the idea that aging creates a ubiquitous risk increase for cancer incidence. Metastasis – moving from the environment where the tumorigenesis initiated – giant hurdle for success but likely to be huge number of cells that slough off. Idea that youth is associated with “Healthy Neighbors” proximal to the initiating cancer cells. Aging is not just accumulation of mutation: idea that the tissue changes create the promiscuous setting. Essentially: the behavior of a single mutation is not equivalent in young and old environments
Tuljapurkar: Response to challenge changes with age – makes the case that the amplitude of the response in young is muted and un-muted and over-amplified in aged, although I would think of it as a disconnection in the response whether that be under or over reactionary. I was very taken with the idea that response to a challenge could push you out of the equilibrium space and into a different state altogether & that you would need to consider the following: Depth of well; curvature of the well; size of the fluctuations.
Spencer: very interesting model for thinking about non-genetic sources of heterogeneity looking at individual cells through the lens of Proliferating v quiescence as a cell state. Spontaneous heterogeneity in asynchronously cycling cells used to identify key nodes in dictating the pace of cell cycle - really nice cell biology and time lapse imaging to uncover CDK2, p21, and stalled forks in the mechanisms. Interesting observation that Mothers pass damage on to the daughters so that the intent to enter quiescence already established in G2 of the mother. Genetic approaches to manipulate CDK2/p21 show that the slow cycling cells have higher stress tolerance – If you force CDK2 in the pausers have a fitness deficit - I wonder though if this is just because cells not ready for division were forced into it creating a vulnerability to stress rather than exposing a beneficial role for the pause.
Natterson-Horowitz: Using examples from different species the relative balance of sympathetic/parasympahtetic v vagal response to stress was explored. High-adrenergic events: eg sudden cardiac, death cardiomyopathy refelctive of a dysregulation of autonomic balanceTonic immobility in response to attach – seen in many different species. Alarm bradycardia is primarily a juvenile response. Not called syncope (because not people) but it sure looks like it. As the animals transition to adulthood they swap over to the sympathetic/parasympathetic response. I do wonder how this is coordinated and communicated with maturation - could there be a signal to indicate that a critical physical threshold had been reached for example, like myokines even? I like the idea that with age there may be a disconnect where there is a loss of the ability to toggle between sympathetic/parasympathetic and vagal responses☂Open questions are whether early experiences inform future balance in terms of response to trauma, Can we build predictive models?
OPEN DISCUSSION: Identified 3 themes: 1. Vulnerability with age, 2. Variance among individuals, 3. Complexity
Some of the ideas that caught my ear include the following: a) The theme of return to homeostasis; b)Changing landscapes with age – and changing landscapes with exposure, the Idea of tipping points applies here too, are there sets of perturbations that would allow you to track their responses and use that to predict the transitions among wells. Expectation that the well tracks with fitness may not be realistic – items roam around the ridges, c) idea that using “Layers” might be a useful way to parse the different components, can we use layers to frame the aging landscape – feedback loops among layers. This concept is likely important for understanding resilience, d) I really liked the analogy to collapse of ancient civilizations: ebbs and flows – tendency toward senescence – more complex, more overheads, more fragility. Is there a way to articulate loss of integrity or resilience and how it might be conceptually related to ecosystems or larger scale entities? so you might think of system failure as an emergent property, e) Nailing down the semantics – framing the language by context would be helpful for developing strategies to move these ideas forward.
Levine: Changes on the molecular level propagate up through the system. I really liked the concept that some degree of failure is tolerated up to a point. Important goal to provide an endophenotype that allows evaluation of intervention efficacy, provide insight into aging - the strategy is to work with risk identification – biomarkers of vulnerability. The methylation "clock" has been developed independently by several groups. Want it to predict more than chronological age so the goal now is to capture the true residual. There is some fascinating biology during development: under age 12y get a boost in methylation that has steeper trajectory from about 15y on the slope is consistent up to 60y. Interestingly, clocks with highest association to chronological age do not have the best predictor ability for disease when age is used as a covariate ☂. Looking now for gene networks – weighted gene correlation networks and then establish associations among networks and the various clocks – amazingly mitochondrial function OxPhos is a primary network ☂. Great next steps will be to take any given module and ask how will they are conserved across the clock – see congruent and discordant depending on what is being looked at for a given context- or in other words, which piece of the clock is driving the associations at different age group and diseases – building “disease-ome” maps.
Schneider: In traditional strategies to determine the resilience of hosts you infect and look at the dynamics of load and symptoms make a regression – generate a tolerance curve. Works at the population level but not at the individual level. To get around this employ a strategy where you generate phase plots – hysteretic curve- this nicely demonstrates that at each point in the infection you ae in a different space- curved around arrow. There was a wonderful theme centered on how to best visualize complex challenge and recovery data - so creative! One idea was to explore Resiliance/Tolerance/Robustness versus Frailty. Taking this a step further you can add new dimensions – generating a disease manifold. Superimposing genetic diversity on this framework revealed the beuatiful underlying biology: Cytokines, metabolome: orotate, parasitic density, granulocytes, reticulocytes. See basically the same changes among the different mouse strains but occurring at temporally distinct phases but always tracking with the course of infection and recovery. In a higher dimension see a wave move through the system – like ripples on a pond. As the shape of the wave changes that indicates where you are in disease course, if the order is perturbed then that might indicate morbidity. This is perfectly set up for exploring aging and comorbidity in the loss of implementation of the infection response and recovery. Some very interesting metabolism is at the heart of this response. Parasite load doesn’t correlate well with outcomes across lines but metabolic status going into the infection does.
Scheffer: Resilience: capacity to recover from challenge, or how quickly homeostasis is recovered after perturbation. Idea of brittleness. System state v conditions (and their interactions)– tipping points where you have two basins in unstable equilibrium – loose resilience when basin of attraction becomes small ☂. An interesting example is abundance of particular bacterial group in the gut, the landscape changes with age and there are apparently tipping elements in the human intestinal ecosystem. As you age more likely to have a particular population distribution. A really fascinating idea is that of Universal rules that govern critical points – true for all dynamic systems where critical slowing down occurs at mathematical bifurcations. The idea is that there may be generic early warning signals for tipping points that could potentially be used for determining indicators of resilience.Using this concept, you might investigate natural fluctuations – the uniformity and small scale of resilient systems becomes increased in variance in the less resilient and so the fluctuation change serves as a dynamic indicator of resilience. The emphasis is on patterns of micro-recovery that inform of the dynamics of the system rather than on the state of the system. If the system as a whole reduces you then get more cross-correlations between subsystems. Failure in one system is communicated to other connected systems. This is a really clever idea and a terrific match for understanding the slippage of system integrity with age.
I
A. A group of few simple species can evolve into a large number number of interdependent species. There is an information theoretic entropy increase in this process. This means that if you reach into the ecosystem and randomly pick a species, you are more uncertain about what you will find that you were initially.
B. There could also be a more direct irreversibility associated with ecosystems: do larger and more complex organisms like us generate more heat (and hence entropy) than simpler organisms? This will need to be answered experimentally. How does entropy production per unit time compare among 100kg of bacteria, 100 kg of insects and a human weighing 100kg? Are complex organisms more efficient at using energy and resources than simpler organisms?
C. '''Jacopo's''' discussion on evolutionary games reminded me of a paradoxical class of games called '''''[[wikipedia:Parrondo's_paradox|Parrondo games.]]''''' These games involve a combination of games that are all losing games, but when played in succession lead to a winning strategy. They have recently been used to explain some ecological and biological features (see references within the link).
'''Pamela:''' Could you please post references and /or tell us about the data pump techniques you used?
'''Fernanda''': I like your philosophical idea of finding interactions where all organisms benefit. The second law does work against us by stating that for order to increase somewhere, there must be disorder created elsewhere. However, I do not agree that two species that mutually benefit must compete with or harm a third species. They could be harnessing energy from abiotic sources such as the sun, wind or thermal vents. Is it mathematically possible to have systems with only positive interactions between the living components? Are there any such systems on earth? +
D
Alfonse Hoekstra's discussion of multiscale resilience was fascinating to me. The network simulation models of Dervis and Peter Hoffman were very interesting and provide useful insights, but to mimic the complexity of human physiology, we would need hierarchically structured networks. I wonder if there are some invariance principles in multi-scale resilience that could reduce the degree of complexity of this type of modeling. The principles and results of hierarchy theory could be relevant here.
Sanne's talk on DIORs (dynamic indicators of resilience) was also quite interesting. There are several open questions here: how to model temporal autocorrelation; how to handle non-stationary time series; how to do systems identification with DIORs, i.e. how can we predict responses of frail/nonfrail using estimates of DIORs. I also think the idea of reactive tuning to stimulus can be examined using novel metrics of DIORs. I would be interested in exploring these ideas in my work!
The second day's talks were also very interesting. Heather's case history was captivating, highlighting the challenges of treating a human being as a complex physiological system. I liked her point that we need to observe and let the system tell us what needs to be done. Ingrid's talk was very informative on the modeling of complex ecological systems. Porter's talk on the resilience of the Pueblo Indian nation to colonization was very educational for me. I can relate to my own Hindu/Indian culture's resilience in having survived several invasions and colonization over the centuries. The idea of axiology, the systems of values which provide the core resilience to a culture, was most interesting. Warren presented some exquisite data on mouse resilience. To me, tlis hehighlighted the huge potential of using mouse models to develop a comprehensive modeling framework for resilience. +
An excellent first day. We heard theory-based perspectives and came directly upon the challenges of human subjects research. The theory based perspectives from Dervis Vural, Peter Hoffmann, and Alfons Hoekstra illustrated the (relative) simplicity of models that effectively abstract and recapitulate several well-recognized characteristics of human aging and frailty. Yet human-derived data are messy, do not lend themselves easily to hypothesis testing because they are so often observational and incomplete, and are confounded by the outbred nature of humans, their varying allostatic loads, and the variety of acute-on-chronic illnesses that bring them to research studies and/or clinical care.
The most interesting part of the second day, perhaps, was the presentation on resiliency among the indigenous peoples of NM. It became quite clear that the ability to (re)generate networks and interactions was foundational to regenerating the population. A reasonable inference , and mirroring Dervis Vural's presentation on Day 1, is that the capacity to reconstruct networks is foundational to resilience. It is unclear whether it is the ability to reconstruct some evolutionarily specified or developmental network is required, or rather a more general capacity. But "fixing a failed node" is unlikely to work unless that failed node is the foundational "network spawner".
The need for a marker that reliably informs clinicians that the capacity to recover from perturbation is now (and forever) exhausted is apparent. The problem around end of life is acknowledging that it is indeed end-of-life, that the physiological derangement exceeds reparative capacity, with or without the stabilization that clinicians can provide. As technical medicine gets better at dealing with minutiae, such markers of inevitable collapse become even more important. As a reminder, 27% of the US Medicare budget is routinely spent in the last year of life, with a substantial uptick in the last month as part of the "rescue phantasy" to use Freud's term. "slowing" and delayed correction of spontaneous or engineered perurbations is a start, but seems by itself insufficient as a basis for clinical decision making.
H
April 9, 2019
Charge for working groups
# Come up with a list of major ideas/problems/concepts that you think we need to work on.
# Think about what conceptual areas could be linked to better address the major questions discussed in #1.
Apr 8, 2019
A few thoughts about general questions for discussion:
# What do we mean by "Biological Failure"? Aging? Senescence?
## Is there such a thing as a truly non-aging organism? An immortal organism?
# Things that change with age...
## Why do so many things appear to increase exponentially, and in parallel on a log-linear scale, with age?
## Are there commonalities across other levels of organization with respect to how things change with age, and by 'things', this could be function, or selection, or failure.
# What maintains variance among populations in aging. Even after controlling for G and E, we still see high levels of variance.
# Issues of complexity/simplicity and networks were touched upon today, but we have not yet gone into detail, discussing this.
Bernie Crespi
Suggestions idea that complexity can be bad. We can think of organismal organization (and function and failure?) in a two-dimensional space of couple (loose<-->tight) and complexity (linear<-->high). Berni suggests (I think) that biological entities with high complexity (brains, immune system) are more prone to bring the entire system down with failure. Entities that are simple and loosely coupled are less likely to fail badly. Suggests a negative correlation between senescence and intelligence, though the GWAS data supporting this are problematic (like the recent study claiming to find genes for SES--https://www.biorxiv.org/content/10.1101/457515v1) are likely due to social stratification. Bernie's neologism of the day: ''badaptation.''
Dario Valenzano
Comparative analysis of annual life history in killifish. Would benefit from Nathan Clark approach to look at rates of protein evolution in annual/perennial species: https://elifesciences.org/articles/25884. Expansion of mitochondrial genome is striking. I wonder if the finding of increase in genome size in the annual species is true of annual plant species (like corn and rice, which have very large genomes) relative to perennial plants.
James deGregori
Notion of adaptive oncogenesis, with stem cells well adapted to niche. As tissue ages, stem cells are no longer well adapted to the niche in which they find themselves. I wonder whether these models are also relevant to the selection process that happens *within* a tumor once cancer growth (and mutator phenotypes) is underway.
Roz Anderson
CR animals show a few major clusters of correlated -omic features, while AL animals show a very large number of small molecules. We should discuss just what these correlations mean, both statistically and biologically, and why these correlation structures (adjacency matrices) differ among the two groups so dramatically.
Tulja
Finishes talk with three general questions:
a. How strong is the trade-off between added longevity and lost fertility
b. Can we explain environmental plasticity?
c. Does stochasticity matter? On any time scale?
For all three of these questions, I wonder if high-dimensional assays (metabolome, epigenome, etc) might have something to add to this discussion...
Sabrina Spencer
The technology that Sabrina is development could add tremendous power to the work now ongoing to track yeast cells as they age in real time. Also shows that quiescent cells are resistant to various stressors. Is that simply that they are metabolically quiescent and so not taking in these toxins?
Barb Natterson-Horowitz
Suggests that the disease-associated heart responses that we see could be maladaptive responses that evolved for adaptive reasons ('capture myopathy', 'alarm bradycardia'). Barb finds evidence for these phenomena in non-human species, but these could well be a large underestimate, simply because the vast majority of these events are never observed, and when observed, not reported.
P
As I am working with conceptualizing the formation of human settlement patterns I came searching for the state of art knowledge on socio-ecological processes influencing their formation. Santa Fe and the Institute greeted me with even lot more than expected starting from eclectic societies of art and science in the city to the bold focus on "the most important" backed by art and mathematics in the SFI.
The course on population and environment offered a wide range of bits and pieces related to demographic dynamics, population size, migration, economy and spatial processes all essentially connected to my own study. All those bits and pieces created a confident methodological backgrond that significantly advances my work. Several ideas in the talks and personal exchange with other participants were directly related to my work. The themes of demographic transitions, migrations, ecosystem services gave me a lot of tought food. More philosophical questions asked in population axiology presented paradoxes in maximization in ranking which in addition to ethical inquiry also makes you think on the level of abstraction of models.
Another topic I was secretly pursuing was the projection of long term processes to the future. As my own work is based on archaeological data from the past, envisioning it towards to the future creates a powerful motivator. And I did observe several interesting trends. Although in most parts we were presented analytical aggregate models the reasoning and understanding behind those macro-level processes involves choice. So a somewhat emotional takeaway from the course for me was - if we want to project from the past to the future, we need to conceptualize our models to the level of choice, and even further to the level of agency (and hopefully use agent-based models on the go).
I really thank SFI, all the organizers and participants for the great opportunity. +
D
As a clinician, it resonated with me to hear Dr. Vural describe that in his models, sometimes "strategic repair" may be necessary in order to re-stabilize a complex system that is progressing toward critical decay (but has not yet reached the critical point). I like the notion that success or failure of the whole system could depend on the order of which nodes are repaired first. I am often faced with the clinical challenge of multi-organ failure and often an intervention that would benefit one organ system might put another at risk, so it is hard to know what sub-system to prioritize. If we could understand the human system better and it could guide "strategic repair," this could have real clinical utility.
It also resonated with me to hear Dr. Hoekstra's description of a similar stressor resulting in vastly different outcomes in his model systems. If there are feature of the system BEFORE or JUST AFTER the stressor that reliably indicate which outcome is going to occur, that would be very useful for prognostication and for making treatment decisions.
Dr. Gijzel's descriptions of data management challenges when dealing with time series human data was helpful (and also cathartic, because I deal with the same challenges). The tension about whether analytical decisions should be guided by conceptual framework and clinical judgment, as opposed to empirical decisions, was something I recognized.
I feel more strongly that the field will benefit from semantic harmonization and precise terminology.
My perspective has changed in that I will be more attentive to opportunities to study resilience changes across the lifespan. I learned about childhood metabolic changes I was not aware of. I learned a new parameter for characterizing temporal autocorrelation across varying lag times. +
H
Bernie Crespi:
The hypothesis/framework presented that risk of failure increases with linearity and tightness of coupling in the functioning of a system is interesting and stimulating, but I wondered about how to quantify these two axes in a practical manner in order to test this idea. How easy is it to classify biological systems on these axes? How much of this classification is artificial? I was also puzzled about how to separate these processes (brain functioning, immune response, and others) given their interactions and integration; but if all these different systems are integrated, then does it make sense to attribute risk of failure to properties of each system? Can this approach be adapted to take into account this integration?
How to define complexity in this context?
The fact that some aspects of senescence may be due to overdefense in later age, seems in line with other observations and theories that senescence may be the result of protective systems dysfunctioning with age; I wondered about the role of our interaction with infectious diseases in shaping our biology and therefore our life history through these mechanisms.
I would like to learn more about how mental disorders match "attractors" potentiated by tight coupling; should read more about this.
Dario Valenzano:
A very nice study system in killifish populations and species where climate/aridity/habitat (temporal ponds) correlates with divergence in life history. Lots of different approaches (QTL, genomics, comparative analysis, phenotypic studies of aging, mutation rates estimates): overall suggest that populations/species living in drier/more ephemeral habitats have accumulated a high genetic load affecting aging and lifespan through relaxed selection, which may be due in part to their demography with frequent bottlenecks. Nice to see such a complete study showcasing how mutation accumulation affects aging evolution. Reminds me of the work by the group of Christoph Haag on daphnia (also in ephemeral ponds)); see reference attached.
James DeGregori:
claims for a change in paradigm in our way of understanding and modelling cancer: number of mutatations would not be so relevant, because of selection on malignant cells. Model of fitness landscape with different peaks, changing with age. What are the mechanisms that make malignant cells unfit in young individuals (niche? policing by other cells? Can we model the interactions between cells to generate such fitness landscapes and their change with age?
What does explain that the increase in rate of cancer on a log scale is linear with time, with the same slope: can we predict that? what does it require in terms of assumptions?
Rozalyn Anderson:
Effect of caloric restriction on lifespan in 40 years long term experiments in monkeys; growth and immune function intertwined. What is the effect of caloric restriction with infectious disease? More generally would a better integration of infectious disease-driven evoluton and life history/aging evolution be necessary?
Shripad Tuljapurkar:
Hidden Markov models driving transition rates could be fitted using longitudinal data on health, frailty and morbidity: could we then estimate how fitness or potential wells change with age?
Barbara Natterson-Horovitz:
Fainting in young individuals broadly distributed taxonomically; could be an adaptation to escape predators in inidviduals that have no other options; links with capture myopathy
What are the life history correlates associated with capture myopathy vulnerability; points to insights in considering relationships with other species (here predation) in shaping evolutionarily features associated with various conditions
Morgan Levine:
aging dynamics at different levels of organization or different aspects of aging (biological, phenotypic, functional) may be different; what are the articulations between these different levels and aspects of aging?
Idea that markers of biological age may be improve our predictive power of condition beyond predictions based on chronoogical age; changes in methylation with age; predictor of chronological age based on methylation age (R=0.98 idea that residuals may inform on condition, but not much residuals variation to play with? not much extra information? methylation not as predictive as biological markers; methylation increases in all tissues but not at the same rate with age: why and how is it related to function?
Schneider & Sheffer: in both presentations, slow recovery signals approaching a critical tipping point; can we get an idea about how close we are from the tipping point? Martin says that there are no obvious way. With systems that have no fixed points as equilibrium but naturally cycle, it is also harder to find signals of a critical transition
general discussion: questions that I like= 1) how to connect ideas about tipping-points/resilience/netowrks to classic evolutionary models of aging, 2) can we design simple models predicting exponential increase in rate of mortality with age? with constant slopes? 3) can fitness landscapes/potentail wells be more than useful images to help us think about aging? Can we parameterize these models? Can they be used to make more quantitative predictions?
other ideas: can we build models with both antagonistic pleiotropy and mutation accumulation and study their interactions? are there distinct predictions about collapses of networks based on these different source of genetic variation affecting aging?
role of social interactions at different levels in biological failure (cell to individual), role of interactions with other species (parasites, predatiors, mutualists)
W
Big idea: consideration of multifactorial definition of and function of sleep
Susan Sara: Hippocampal ripples are important for memory consolidation; how would the timing of increased ripple density post-learning change with species and/or development? Can this be connected to brain size/networks involved in memory consolidation?
Gina Poe: How are memories tagged for strengthening or weakening? What computational approaches could inform our understanding of this? Complexity of memory likely influences the time scale - how might we explore this experimentally or computationally?
Sara Aton: Mechanism whereby a rate code is translated to a phase code. How does modulation of neurotransmitter milieu affect the propensity of network to propagate particular rhythms? How does the power of rhythms relate to metabolic rate of the brain? Does synchrony play a role?
Kimberley Whitehead: How do mechanisms for spindle-burst production relate to mechanisms for spindle production? What are appropriate comparisons for pre-term human infants and animal models of different species (rat, sheep, etc.)?
Beth Klerman: Seasonality is observed in the duration of the biological night as measured with melatonin expression; how does this relate to mechanisms for seasonality encoded in the SCN? Important new direction to take insights from groups and translate this to predictions for individuals? How do we handle missing data and investigate the tolerance of models to missing data?
Victoria Booth: What do changes in the structure of sleep tell us about changes in physiology? Blumberg showed a lack of consolidation of wake in OXKO infant mice emerged during development when WT mice were able to consolidate wake bouts to transition to a power-law-like behavior. We showed similar differences in a mouse model with acute orexin cell loss (Branch et al., SLEEP, 2016). What can we learn about sleep in development in the absence of orexin neurons? How would physical distance between neuronal populations and degree of myelination affect the time scales on which these populations interact? For example, would slower transmission allow more bistability between states? Neonates have more wake to REM sleep transitions. How would we understand this in terms of different network configurations for NREM/REM regulation? Would this be a useful constraint?
Jerry Siegel: Fur seals can switch between bihemispheric sleep and unihemispheric sleep. Modeling has suggested that this can be understood based on relative strengths of contralateral and ipsilateral connections between brain hemispheres (Kedziora et al., J of Theoretical Biology, 2012). How do predictions compare to anatomical evidence? What other insights from physiology could we get from this system? Important to consider the role of temperature and thermoregulation in sleep dynamics. This has been explored a bit in a recent model (Banuelos et al., Effects of Thermoregulation on Human Sleep Patterns: A Mathematical Model of Sleep–Wake Cycles with REM–NREM Subcircuit, in [https://link.springer.com/book/10.1007/978-1-4939-2782-1 Applications of Dynamical Systems in Biology and Medicine] pp 123-147). What are other ways that these ideas could be included in mathematical models of sleep?
Van Savage: What can theory tell us about the quantities we should be measuring experimentally? For example, the theory suggests the importance of considering the ratio of time asleep to time awake rather than durations only.
Geoffrey West: Developing a theory that connects sleep duration and lifespan. Lifespan scaling is derived from ideas of sleep for repair; could we introduce a modification that involves a different role for sleep in development? How is this impacted by differences in species development/maturity? Can we understand this scaling in terms of the DNA methylation view of aging (e.g., converting dog years to human years: https://www.sciencemag.org/news/2019/11/here-s-better-way-convert-dog-years-human-years-scientists-say)?
Alex Herman: If synaptic density drives metabolic rate of the brain, is there also a role for different rhythms? Reporting a sharp transition in the relationship between the ratio of time asleep/time awake to brain mass. What else does this transition correlate with? Changes in behavior? What would this be abrupt and not graduate? What does synaptogenesis peak and how does it relate to this transition? Follow up with Junwu Can regarding her method for threshold detection of this transition.
Bob Stickgold: Different types of sleep for different types of learning; gist words - how does this relate to other measures of creative thinking? What is the role of forgetting for dreams and what is important about the dreams we remember? What about problem solving that can occur while sleeping/dreaming?
C
Brains fail - like bridges, like suspenders, and like relationships. The brain constantly operates at a point of criticality - at least for our optimal cognitive state. Adaptive systems with connectivity often have cascading failures. Can the brain heal itself? Is this robustness? Can that self-organization go wrong, or be improperly applied? It's not a bug - it's a feature. I am reminded of the "swiss-cheese model" familiar in root cause analysis. Multiple failures must be serially associated
Why do we care about a diagnosis? What is meant by a "proper" diagnosis? Does "diagnosis" imply stationarity. Must we have a tight mechanism to have a diagnosis? Or perhaps simply a cluster of symptoms. Or a basis in which to guide therapy I approach my patients based on what is the next thing I am going to do. Sledgehammer solutions. - may be best. Borrowing from vaccines, can we look at treatments as "learning". Pain is an example for a top-down approach to disease. Much like traditional Chinese medicine.
Highlight 1
Richard's buzzing that described experimental determination of the boson and how it related to inconsistencies in Alzheimer's Disease was simultaneously the most confusing and the most entertaining part of the conference. +
W
Common principles of sleep-related oscillations that subserve some form of 'learning', e.g. bursting on a background of neural quiescence - this definition would encompass the unusual features of sleep in early development. Very interested by a) Susan's reference to the Aston-Jones 1981 paper that shows attenuation immediately prior to firing as if to 'boost' the effect of the firing, b) that oscillations can occur but the ones that have many frequency bands coupled together are the most important for memory consolidation, c) Susan's quote 'synchronous activity during sleep provides a substrate for []' (the square brackets could be replaced by nearly all functions!)
Remembering: we need augmented oscillatory power for a stimulus to be learned (Sara's talk) and this can be artifically provided (Ognjanovski et al. 2018), how might this relate to augmented sleep EEG power post brain injury - is it possibly not just a biomarker of damage but a marker of the synchronised activity necessary to 're-learn' new circuits? If a neuron which has learnt something continues to stay more active, for how long? How does that link to the renormalisation mentioned in final slide?
Forgetting: Is it REM sleep or forgetting or dreaming specifically for forgetting? Very interested by Gina's reference to Crick 1983 Nature paper - how could that relate to infants who have MORE REM sleep but probably don't yet dream like we conceive dreaming
Sleep oscillations as a spatial filter: because some neuronal ensembles can't keep up with the speed of oscillations, creates a spatial filter
Cortical region specific projections from Locus Coeruleus (Chander et al. 2019 ref from Gina's talk): a way that sleep-wake states could differentially modulate sensorimotor areas in the developing brain, to help them to differentiate?
Durations of sleep: Bob's conceptual theory that 100ms of sleep would be sufficient to encode a memory e.g. a microsleep. Could that be reversed to think about some adaptive function of the extra-short WAKE durations pre-term infants have e.g. 2 minutes?
Body weight vs. age predicting sleep-wake bout durations: Following up on Geoffrey's question, I will go back to the literature to check those references that biological age is more predictive than body weight. But thinking about his question, my forthcoming project on intra-uterine growth restriction would be an ideal opportunity to look at this in more detail.
Other ways I intend to improve my practice following this meeting: report ambient temperature
Evocative metaphors!: 'Mountain of wakefulness'. Bob's 'Memory evolution', more dynamic than just consolidation which doesn't fully capture sleep's functions. 'Ground truth' - the challenges of establishing this in many aspects of sleep research
Does LC firing trigger upstate or is LC firing triggered by cortical down state?
Can paramaterizing models with anatomical distances and scaled time-constants lead to emergent behavior in dynamics of sleep wake cycles consistent with data?
What is the history dependence of sleep states? Are all REM and NREM sleep bouts created equal over the course of a night?
What
Oscillations: What sets the intrinsic rhythm of a cortical area? What reads out the phase-ordered temporal sequence of spikes?
Can animals have more localized forms of sleep?
How are the electrophysiological signals of REM sleep related to its functions?
How can animals with close phylogenetic relations show such different sleep patterns?
Do we have a phylogenetic history of the different functions/phenotypes of sleep? +
D
During the first day, although the talks were very different, they very nicely complemented each other. It is amazing to see that we are all adopting slightly different approaches to investigating resilience in human aging but that they can all be placed in the larger framework of resilience of complex dynamical systems. We are all pioneering in this area and only sharing our struggles and recent insights was already very valuable, at least in my experience.
During the second day, I began seeing that we are working along 2 parallel lines:
#Finding ways to quantify resilience / resiliencies
#Increasing our understanding of the dynamics of the complex system in terms of resilience
Some reflections:
- I liked Alfons' idea of making real-life examples of "Resilience is ........" in the form of a short narrative / artwork / graphical illustration / equations. I agree that these can be very helpful to increase our understanding of resilience and involve more people (e.g. clinicians) in the resilience thinking and discourse.
- Marcel commented that for humans, it's clear that there are alternative stable states in health, but we do not know what are the precise perturbations and positive feedbacks causing the transition. To increase our understanding about this, I think we need to start with making mechanistic models. We can use these mechanistic models to generate new hypotheses.
- Ingrid pointed out the difference between acute stressors (perturbations, e.g. a stimulus-response test) and slow stressors (drivers, e.g. aging). +
C
Excellent presentations were given highlighting the descriptive power of convolutional deep networks, also illustrating its partial explanatory power and where it fails. This pointed out some interesting ways forward that have to go beyond their current architecture, in particular taking dynamics into account. Interplay between structural and functional connectivity was highlighted. Limitations of stationary metrics were evident (functional connectivity), but nicely showed how far it can be pushed successfully in applications. Model approaches providing explanatory approaches were often too simplistic, not in terms of realism, but in terms of simplifications of concepts (brain states, behavior, as static entities). In the discussions it was evident that there is a need for a formalisation of the internal state dynamics of the brain, before perturbations can be applied to it (breaking the brain). A formal frame work for provision of and recovery from such perturbations is needed, several good attempts were provided and need to be pursued in the future, and supported by data. Need for individual predictive capacity of these frameworks was highlighted rightfully. +
A
Experimental models to study cellular aging vary dramatically.
We clearly need a consensus definition of aging, or more specific concepts. Is aging the loss of specific functions, the loss in the ability to divide, "senescence" (which itself does not have a consensus definition), the movement towards mortality, or the accumulation of "information" over time? Can there be a single definition of aging across the tree of life - from single cells to complex multicellular organisms like mammals?
If the definition is a functional one - aging is the loss in the ability to perform X function, then aging needs to be contextualized. Organisms at different scales (prokaryotes vs birds vs humans) have dramatically different "purpose" in the living world, and they carry out very different functions. Is there one type of aging that unites them all? Or qualitatively different forms of aging, or aging processes?
An interdependent challenge with the previous one is the issue of measurement. What are good measures of aging - again it depends on how it is defined.
If aging is defined as ''something'' that tells us how close to death an organism is to end of life (i.e,, mortality) or to loss of function, then it implies that aging biomarkers need to be developed prospectively. In other words, the aging marker need to predict some future behavior. One example is the DNA methylation or epigenetic clocks. +
D
Fascinating, multi-level, presentations and discussions.
Some points that have come to mind for me. Concerning Dervis' model, I was thinking how having the nodes have an internal structure, that is, each node would have a network structure -- self-similar structure -- would affect the dynamics... I think such an approach would connect nicely to the other presentations concerned with the multi-scale organization of the system.
Concerning the analysis of physiologic signals and the ability to extract different types of information, the site https://physionet.org contains a lot of useful information and may even include a community for discussion of issues (sadly, I haven't worked in this area for a while now).
Ingrid's talk has raised many interesting questions about the different types of models and their different purposes. My thought is that we need research occurring at all types of models with the understanding that each type can contribute to the others. For example, the parametrized, complex climate/ecological models enable us to conduct computational experiments that otherwise would be impossible. However, because of the complexity of the model it is impossible to gain insight on which specific experiments to conduct. Simpler, stylized models, that could be developed and tested against the complex models, could provide the insight to select what computational experiments to conduct.
Axiology is an extraordinarily interesting concept. Knowledge is a responsibility not a right for the Pueblo peoples. When we talk about resilience, what are the things we are valuing... Resilience of Pueblo peoples to colonial injuries: system had developed redundant relations that can prevent failure in case of injury to system.
Porter's talk reminded me of a study of mortality in Chicago some decades ago during a particularly extreme heat wave. Tow communities with similar socio-economic, ethnic and educational characteristics had very different outcomes -- the strength of the social networks was the aspect that distinguished the communities and the one that predicted the outcomes.
P
First and foremost, I would like to thank the Santa Fe Institute for this incredible opportunity! This experience was intense and intellectually stimulating in many different ways. As an archaeologist, I found Dr. Lee's presentation on agroecological and environmental-dependent demographic models to be fascinating, and I certainly can see the relevancy of such models in generating better prehistoric population estimates. In addition, I found Dr. Hooper's presentation on statistical model selection to be very useful. Last, but certainly not least, I truly enjoyed the opportunity to interact with my fellow attendees. There is an incredible amount of value in simply chatting with so many interesting and intelligent people in one place. In my opinion, the synergies that derive from the aggregation of so many great minds in one location is one of the many benefits of a place like SFI. I hope to have the opportunity to return to SFI in the near future. +
D
First day: I learned about modeling strategies that can be used to better understand potentially universal properties of damage and repair of the dynamic system. Some were compared to empirical data. Hormesis was introduced (in the context of bone health) as an important consideration in modeling resilience. Open questions included that some patterns/observations obtained from simulations remain to be explored further (e.g., three 'trajectories': die, recover/die, & recover). Still open for me is how to actually apply some of the great ideas discussed today. For example, I had already considered the life-course in outcomes in (older) adults, but still don't have a good handle on how to actually incorporate or study this in an already aged population, or if it's possible. I have some of the same questions regarding the most useful (pre-)processing of time-series data. However, the 'middle out' approach seems to be a useful way to reduce the dimensions of complexity associated with modeling.
Second day: The extended discussion on the differences in the definition of resilience (e.g., engineering vs. ecological) and the addition of the concept of reserve further highlighted the need for standardization of definitions to make sure researchers are all on the same page. Mechanistic models and mice models show promise of better understanding the dynamics of the (aging) human, but caution is advised in trying to translate these interpretations. A case study brilliantly demonstrated that apply these concepts to 'real life' situations (e.g., patient care) requires much more thought. This was reinforced by another case that was interesting, not only from a network interaction standpoint, but also because the patient often knows their state/potential outcomes better than typical 'objective' tests.
Particularly impactful for me was the example of resilience on a community level in the Pueblo people. I find it a wonderful model to follow for understanding what factors contribute to the resilience in other contexts.
