Kevin
Cho
Mechanistic Basis of Cognitive Rigidity Due to Early Life Stress in Females
Abstract profile. Full document pending author claim.
Authors:
Kevin Cho, Saad Hannan, Takao Hensch
Date Created:
2025-01-01
Course Title:
Professor:
Not specified
About Paper:
Adversityinearlychildhoodresultsinsignificantconsequencesfor immunohistochemistry and confocal imaging, we focused on the development, and is associated with increased risk of psychiatric anterior cingulate cortex (ACC) as a region implicated in rule disorders. Females are particularly vulnerable to stress-associatedlearning to analyze the formation of perineuronal nets (PNNs): disorders of depression, anxiety, and cognitive impairment, but specialized extracellular structures found primarily around PV the mechanisms leading to these sex-based differences in response neuronstopromotetheirmaturationandclosesynapticplasticityin to early life stress (ELS) are largely unknown. Differential early development. We compared arborization of PNNs between effects of ELS have been reported in mouse models, with female ELS and control care mice to probe the maturity of PNNs in mice exhibiting resistance to attentional deficits following ELS, females at an earlier timepoint than physiologically expected as a yet difficulty with rule reversal learning, a sign of cognitive potential early biomarker of ELS and resulting cognitive rigidity. inflexibility. Previous research suggests that parvalbumin- Future research will involve utilizing fiber photometry to record expressing (PV) interneurons, a major class of inhibitory neurons, synchronized firing of PV neurons during a rule reversal learning may be involved in the underlying circuit mechanisms. Thus, this tasktofurtherelucidatetheirfunctioninrelationtoELS.Thiswork project seeks to closely study the early developmental events that may lead to a greater fundamental understanding of the effects of lead to cognitive rigidity in female mice exposed to ELS. Using ELS in development to inform future targeted therapeutics. Developing a Machine Learning Force Field for Coarse-Grained Monte Carlo Protein Unfolding Simulations Addison Crider, Kibum Park, Eugene Shakhnovich Harvard College | Leverett House | Chemical and Physical Biology | 2028 Molecular simulations, including Monte Carlo methods and Inthiswork,weaimtodevelopacoarse-grainedNNPattheresidue molecular dynamics, provide valuable insights into the biophysical level that integrates with our Monte Carlo Protein Unfolding properties of proteins, which can be challenging to infer from simulation package. We trained the model on three quantum experimentation. However, their computational demands make datasets. To learn non-bonded interaction energies, which are them infeasible when attempting to simulate larger time scales. the most complex, we used a subset of FMO-SCOP-29Jun2022, Thatisacriticalissuewhenstudyingproteinfolding,assimulations containing 4,877 proteins with lengths ranging from 7 to 1,526 often fail to capture large conformational changes that occur over residues. For the more detailed bonded interaction energies, we extended periods of time. used QM9 and SPICE, which include small peptides suitable for capturinglocalchemicalinteractions. BuildingonTorchMD-Net’s To address this limitation, system coarse graining is often used, Graph Network, the model architecture comprises a transformer to which reduces a simulation’s degrees of freedom to improve capture long-range interactions, with ongoing tuning to enhance efficiency, but at the cost of accuracy. Recently, machine learning models have been used as an alternative to classical physics- performance. After obtaining a functional model, we will validate based force fields. These neural network potentials (NNPs) it by simulating a commonly used set of 12 fast-folding proteins and compare the results to those using physics-based force fields. offer the possibility for improved accuracy without a substantial performance loss. By combining both strategies, coarse-grained With the advantages of coarse-grained NNPs, our model will be NNPshavebeenshowntoimproveperformancewhilemaintaining used to study protein folding dynamics, including co-translational accuracy. and chaperone-mediated protein folding, whose timescales often exceed the reach of all-atom simulations.
Abstract:
Adversityinearlychildhoodresultsinsignificantconsequencesfor immunohistochemistry and confocal imaging, we focused on the development, and is associated with increased risk of psychiatric anterior cingulate cortex (ACC) as a region implicated in rule disorders. Females are particularly vulnerable to stress-associatedlearning to analyze the formation of perineuronal nets (PNNs): disorders of depression, anxiety, and cognitive impairment, but specialized extracellular structures found primarily around PV the mechanisms leading to these sex-based differences in response neuronstopromotetheirmaturationandclosesynapticplasticityin to early life stress (ELS) are largely unknown. Differential early development. We compared arborization of PNNs between effects of ELS have been reported in mouse models, with female ELS and control care mice to probe the maturity of PNNs in mice exhibiting resistance to attentional deficits following ELS, females at an earlier timepoint than physiologically expected as a yet difficulty with rule reversal learning, a sign of cognitive potential early biomarker of ELS and resulting cognitive rigidity. inflexibility. Previous research suggests that parvalbumin- Future research will involve utilizing fiber photometry to record expressing (PV) interneurons, a major class of inhibitory neurons, synchronized firing of PV neurons during a rule reversal learning may be involved in the underlying circuit mechanisms. Thus, this tasktofurtherelucidatetheirfunctioninrelationtoELS.Thiswork project seeks to closely study the early developmental events that may lead to a greater fundamental understanding of the effects of lead to cognitive rigidity in female mice exposed to ELS. Using ELS in development to inform future targeted therapeutics. Developing a Machine Learning Force Field for Coarse-Grained Monte Carlo Protein Unfolding Simulations Addison Crider, Kibum Park, Eugene Shakhnovich Harvard College | Leverett House | Chemical and Physical Biology | 2028 Molecular simulations, including Monte Carlo methods and Inthiswork,weaimtodevelopacoarse-grainedNNPattheresidue molecular dynamics, provide valuable insights into the biophysical level that integrates with our Monte Carlo Protein Unfolding properties of proteins, which can be challenging to infer from simulation package. We trained the model on three quantum experimentation. However, their computational demands make datasets. To learn non-bonded interaction energies, which are them infeasible when attempting to simulate larger time scales. the most complex, we used a subset of FMO-SCOP-29Jun2022, Thatisacriticalissuewhenstudyingproteinfolding,assimulations containing 4,877 proteins with lengths ranging from 7 to 1,526 often fail to capture large conformational changes that occur over residues. For the more detailed bonded interaction energies, we extended periods of time. used QM9 and SPICE, which include small peptides suitable for capturinglocalchemicalinteractions. BuildingonTorchMD-Net’s To address this limitation, system coarse graining is often used, Graph Network, the model architecture comprises a transformer to which reduces a simulation’s degrees of freedom to improve capture long-range interactions, with ongoing tuning to enhance efficiency, but at the cost of accuracy. Recently, machine learning models have been used as an alternative to classical physics- performance. After obtaining a functional model, we will validate based force fields. These neural network potentials (NNPs) it by simulating a commonly used set of 12 fast-folding proteins and compare the results to those using physics-based force fields. offer the possibility for improved accuracy without a substantial performance loss. By combining both strategies, coarse-grained With the advantages of coarse-grained NNPs, our model will be NNPshavebeenshowntoimproveperformancewhilemaintaining used to study protein folding dynamics, including co-translational accuracy. and chaperone-mediated protein folding, whose timescales often exceed the reach of all-atom simulations.
Source:
Harvard / Alice Chen, Stephanie Sendker, Rizwan Romee / 2025
Topics:
els, protein, model, used, cognitive, early, female, simulation, learning, coarse, rigidity, stres
