Organized Framework Of And Introduction Into Neuroscience Concepts

Introduction

The neuroscience of intelligence is incredibly complex, no doubt about that. I saw that many newcomers to the community may feel overwhelmed by the amount of information. To help them build a mental organization to contextualize various research papers and findings, I devised a new and improved framework below(more detail in comments by scale category), much more organized and clear compared to sources-discussion:

This framework is divided by scales of measurement, computational complexity, anatomical resolution, and depth of analyses. The literature investigating intelligence is extremely wide in scale, from examinations of micro-scale molecular mechanisms underlying key processes in memory to fMRIs and EEG data correlating higher intelligence to macro-scale activity measures. Thus, I find it intuitive to subdivide our understanding of intelligence by scales most commonly utilized by studies, and then seek to make connections between these scales afterwards. I will list major keywords (use CTRL+F for search) and concepts below for each category to help people to better conceptualize intelligence in a cohesive and less strictly segmented manner. The goal is a smooth and continuous, solid foundation.

1. Molecular mechanisms

2. Cellular processes and morphologies

3. Neuronal networks, small anatomical features, and network computational analyses

4. Brain regions and circuits, large anatomical features, metabolic and broad measures of activity

5. Cognitive abilities, behavioral analyses, psychometric testing

Every comprehensive and workable theory should make sense at each level imo. At the end, I will provide an example of how I utilize this.

1. Molecular mechanisms

These are individual chemical and molecular reactions and pathways. Some examples: negative allosteric modulation of PDE4D long isoforms, positive allosteric modulation of AMPA receptors by TAK-653, NMDA receptor binding of glutamate and the co-agonist mechanism of glycine, etc. Because there are so many of them, and the networks of reaction pathways are so complex, reading about one or two molecules, reactions, or pathways can feel incredibly isolated and leave you questioning why there is research on it in the first place.

For a proper understanding of the significance of each pathway, one usually has to do their due diligence researching the upstream factors and downstream impacts of a certain reaction or mechanism, as well as prominent cell types, expression distributions of key involved genes, etc. I highly recommend learning at this scale by focusing on one keyword at a time. First understand the specifics of the singular mechanism/molecule you're interested in, like receptor dynamics, half life, modulation of reaction mechanism, etc. Then look into upstream and downstream factors, canonical pathways they are involved in, and then move on to broader, higher scale/level understandings such as cell type-specific gene expression distributions, expression distributions across cell morphological features, pathway-based gene ontology, etc.

Here are some keywords that I personally think are particularly important to look into regarding the molecular mechanisms implicated in intelligence:

NMDA, AMPA, glutamate, PDE4D, dopamine, D1 receptor, D2 receptor, MAP2, MAPK, ERK1/2, CCR5, GCPII, mGlur3, mGlur5, NAAG, acetylcholine, MAP4343, nAChr a7, M1 AChr, NR2B subunit, sigma1 receptor, mu opioid receptor, BDNF, IGF-1, EGF, etc.

There are plenty of interesting resources online for looking into specific pathways, drugs, mechanisms, genes, etc (will add more as I remember/find):

• Drugbank
• The Gene Ontology Resource
• HumanBase
• String-DB

@arche note -> HumanBase (and String-DB) links can be found on gene pages when you click the "Search PubChem" button.

Also, if you want a quick and basic (but potentially faulty) rundown, GPT-4 and Claude are decent as well.

2. Cellular processes and morphologies

These can be seen as one step above individual molecular mechanisms; they can be seen as entire molecular pathways, possibly multiple. In research, there are specific terms meant to encompass individual canonical cellular pathways, but in reality, most pathways interrelate and influence each other through a multitude of ways. For the sake of simplicity, this category ranges in scale from individual pathways to more general processes within a cell, and consequently, may or may not impact the cell's individual structure and function.

You can learn at this level by first looking up the keyword of interest, looking into the smaller scale pathways and molecular mechanisms underlying it, then looking at how it impacts higher level multicellular processes and perhaps even larger scale phenomena. But generally, you don't want to skip steps in between scales because that break in continuity makes your understanding more segmented/unintuitively reliant on rote memory.

Ex: dendrite elongation (scale 2) is mediated/determined by mechanisms primarily involving MAP2, MAPK, ERK1/2. Namely MAP2 dephosphorylation biases for elongation of dendrites as opposed to branching, and phosphorylation sites can differ between CaMKII, PKC, PKA, etc (scale 1). It can be impacted by mechanisms/pathways involving NMDA activity, PDE4D long isoforms, pregnenolone, BDNF, IGF-1, EGF, and much more (scale 2). Dendrite elongation, at larger scales, may impact network dynamics, potentially improving neural computational efficiency and cross-columnar connectivity, and may impact the neuron's structural proximity to the ideal fractal (scale 3, then 2).

Keywords (too many to list): dendrite elongation, dendrite branching, distal dendritic synaptogenesis, proximal dendritic synaptogenesis, microtubule mobilization, gene upregulation, gene downregulation, synaptic transmission, dendritic Ca2+ spikes, action potential kinetics, neuronal differentiation, mitochondrial biogenesis, oxidative stress, metabolic efficiency, fractal-ness of a neuron's dendrites, synaptic distributions across the dendrite of a neuron.

3. Neuronal networks, small anatomical features, and network computational analyses

These deal with multicellular phenomena, are consisted of aggregated and complex interplays of many cellular phenomena, and are thus determined by the even more complex interplays of very many molecular mechanisms. At this level, we can finally start to see the computational efficiency of a network and speculate about the relative computational value/role of individual neurons, etc etc. We can also observe differences between networks in the number/density of neurons, the number/density of synapses, the relative rate of firing, etc. More research articles addressing this level of neuroscience tend to not go too deep into individual molecular mechanisms, as they tend to be more interested in computational simulations and larger level measures like the metabolic cost of the network etc. But they often refer to scale 2 phenomena to explain their findings.

Ex: Dendrite elongation as opposed to branching(scale 2) of the key associative role-neurons like L3C neurons in cortical layers 2/3(scale 3), might improve the complexity of information that can be reliably and easily encoded by that finite network, without hindering its metabolic efficiency or causing network hyper-excitation (scale 3).

Keywords: network computational efficiency, network hyper-excitation, metabolic cost, network representational complexity, sparsely active network states, average number of distal dendritic inputs across neurons in the network, variation/deviation of that same metric, average and opposite extremes of synaptic distributions across a dendrite, network synchrony, output complexity and flexibility, cortical columns, within column dynamics, cortical columnar recursive signaling, columnar representations of temporally sequential information, cortical layers 2/3, layer 4, layer 5, layer 6, autism and neuronal hyperexcitation, autism and bias towards local cortical connectivity at the cost of long distance connectivity, autism and columnar size

4. Brain regions and circuits, large anatomical features, metabolic and broad measures of activity

Scale 3 can be seen as groups of neurons to form networks. Scale 4 can be seen as groups of networks to form larger networks, or entire brain regions, and even larger networks, to form entire brain circuits. Research papers at this level tend to be highly unspecific in terms of experimenting with or examining the nuances within each individual neuron. They tend to be a lot more focused on larger scale findings like whether or not a certain region is more or less active in response to certain perturbations, and how a region's degree of involvement in a memory changes over time. At this scale, we also see papers investigating stuff like gray matter volume correlations with IQ, fMRI recordings quantifying cognitive flexibility brain entropy, EEG measures of brainwaves like gamma-theta synchrony, fronto-parietal cross-activation, brain regions most activated in inductive vs deductive reasoning, etc etc. Metabolic consumption of entire brain regions, blood flow, etc, are also included here.

Learning deeply from this level can be a bit confusing, although it is certainly the easiest to start. This is because this is where cognitive science meets neuroscience, due to larger scale measurement tools being easier to develop and use compared to more high-resolution tools. But for a detailed understanding past oversimplified statements and generalizations, we need those higher-resolution tools and datasets. The vast majority of people in the brain training sphere tend to theorize based mostly on research at this level at the deepest. Few rarely even look at level 3. On the other hand, most people interested in nootropics only look at level 1 and level 2. We want to ideally look at all.

Keywords (other than those already stated): medial temporal gyrus, intraparietal sulcus, white mater architecture correlations to IQ, dlPFC, IPS-visual perception networks, resting state fMRI functional connectivity, cortico-thalamo and thalamo-cortical projections, spatial parietal vs visual temporal processing dichotomy, autism vs dyslexia dichotomy, etc.

@arche note -> Subjective factors like APFs and Fracs may fall in this category.

5. Cognitive abilities, behavioral analyses, psychometric testing

At this level, the reseearch isn't too interested in the physiological details and nuances of the brain. More research papers here are interested in statistically teasing out likely and unlikely correlations between various cognitive/psychometric abilities, more so than finding a true basis of certain abilities. In other words, identifying abilities likely to have a physiological basis without proposing how exactly physiologically, and seeking methods which experimentally somehow work to alter such abilities. In reality, more and more papers are at level 4, and have some attributes of level 5 or utilize publications dealing with level 5 to bolster their hypotheses.

Ex: Jaeggi's 2008 study doesn't really investigate too much regarding the physiological basis of improvement, just claiming there exists something like it to explain the results.

Keywords: honestly imo, the terms in cognitive science at this level are unintuitive and often based on historical usage rather than what is sensible in terms of physiological backing. It's very, very difficult to accurately and comprehensively link concrete details from level 4 to level 5 even. But level 5 has immense value despite its flaws, that it can help direct research at levels 3 and 4, by finding what areas of investigation are likely to lead to concrete and valid results at the physiological level, and vice versa. Level 5 is crucial to link physiology to actual human behavior and our conception of intelligence. Ex: WMC and WMM/reasoning (level 5) were validated by physiological data(level 4) which statistically distinguished similar but distinct networks for each function. This in turn, for us, has implications at the level of psychometric testing and cognitive training (level 5).

Examples

A full example:

Dendrite elongation, mediated by MAP2, MAPK, ERK 1/2 (I'm sure there are more but these are some main ones) and their roles in microtubule dynamics, is modulated by various pathways and mechanisms, some of which can be impacted by PDE4D long isoforms, NMDA activity, pregnenolone analogs, and many, many more. Based on computational analyses by Jeff Hawkins and colleagues, studies correlating increased dendritic size but not dendritic and synaptic volumetric density with higher IQ, and the numerous studies supporting the importance of dlPFC, temporal regions, and IPS in crucial functions important in intelligence: we might theorize that specifically in these regions, for key neuronal populations in cortical layers 2/3, especially pyramidal subtypes most crucial for cross columnar connectivity like L3C neurons, increasing dendritic length might improve generalizable cognitive functions. At the larger level, this can computationally translate to enhanced network activity complexity and ranges of dynamic extremes. Specific examples supporting this viewpoint:

• Oroxylin A and molecular mechanisms involving ERK 1/2 and causing dendrite elongation specifically, whereas BDNF actually hinders this slightly to bias for branching, improvements in memory
• Rat training studies showing key regions having increased dendrite length but decreased dendrite density after visuospatial working memory training
• PDE4D and molecular mechanisms, impact on neurite elongation/structure, localization in pyramidal neurons of dlPFC, dramatically improved memory in several animal models including primates.
• Pregnenolone impact on neurite elongation/structure, impact on various neurotransmitter receptor expressions across important regions such as striatum, hippocampus, and cortex, improved outcomes in many disease models and actual patients.
• The association between D1r and NMDAr, and D2 and NMDA activity, reflecting crucial mechanisms of how dynamic extremes of information integration vs filtering are achieved through D1 and D2 balancing dynamics, and the additional role of NMDA in shaping dendritic structures, as well as the sensitization of D1rs by working memory training. More correlational data on intelligence, creativity, brain entropy, etc. More data supporting link between striatal dopamine and working memory training transfer, and frontal dopamine and WMC, etc.

Largely, the above is a cohesive story from level 1 to level 4. But it lacks in level 5, which is why our application of this knowledge to the art and science of brain training itself feels quite limited, beyond what I have already posted in my big theory forum post. Hopefully in the future, more research will more cohesively and comprehensively link all of these levels together.

Learning a new complex topic is hard. But often times, the most difficult part is not knowing what you don't know. I hope the above will help newcomers obtain a broad perspective on what levels of neuroscience can teach them about intelligence, and what kind of keywords, datasets, and insights are featured at each level of scale. Cheers.