in biological systems. This section will briefly discuss the neuron-to-neuron communication process from the neurotransmitters to the synapses in the first neuron, to the interneuron signaling, and finally to the receiving neuron receptors. Neuroscience is an active field, and the details are not fully understood, so it is harder to characterize than the artificial systems counterparts; some perspectives on this are given below. In biological neural networks, there are far more synaptic connections than neurons. Ielmini notes that the synapse-to-neuron ratio is 10,000 (104). This is a considerable number of synapses compared to neurons, which can make crossbar systems quite large to achieve this enormous number (Ielmini 2018). When considering the materials used in biological neurons and neural networks, addressing a few questions can help guide how best to compare the materials layer in biology versus electronics. First, what elements, molecules, biomolecules, and biochemical materials are used for each part of the neuron? There is active research in neuron molecular biology, but there are limited opportunities to access human neurons ethically. Some proteins have been identified to be part of neuronal processing (Davies 2006). Second, what are the critical biomolecular properties that motivate their utilization at the compound level, nucleotide level, protein level, cell circuit level, and nerve cell level? Exploring this question will be one of the themes explored as each architectural level is discussed. The neuron-to-neuron communication process starts from the neurotransmitters to the synapses in the first neuron, to the interneuron signaling, and finally to the receiving neuron receptors. Davies, and his chapter scholars, explore the molecular biology of neurons from various active research points of view. In greater detail, it proceeds with these steps: First, the transmission of the information starts with a discussion of neurotransmitters. The neurotransmitter released from one nerve cell binds to the receptor of another nerve cell, which results in depolarization or other effects in the target nerve cell. A neuron can transfer a signal to a postsynaptic neuron by releasing certain chemicals (neurochemicals) into the synapse and activating postsynaptic receptors. There are many neurochemicals, on the order of 100. Measurement testing has shown that neurotransmission membrane potential changes at certain defined discrete levels, or quanta, must occur in multiples of quanta (Davies 2006). Second, synapse transmission is primarily controlled by neurotransmitter activity. The synapse physically connects with the receiving neuron, and several proteins are involved in this connection and exchange of information. The postsynaptic density is where the synapse connects with a receiving neuron’s dendrite. There are four types of proteins present in this area: (1) plasma membrane, (2) signaling, (3) cytoskeletal, and (4) linker. They are signaling results by forming protein complexes that respond to signals from the membrane surface. Some signaling machines have been identified, which reach the next full compute stack architectural level (Davies 2006). Third, interneuron signaling utilizes signaling machines to transmit the encoded information from one neuron to another. One form of signaling is done through phosphorylation. The fact that there are several types of signaling pathways illustrates the complexity and connectivity that exists in neural networks. There is also a link between the nucleus and signaling. Calcium ions can act as messengers to link the synapses to the nucleus to pass signaling information (Davies 2006). Fourth, signal reception is accomplished by using signal receptors. Two types are ligand-gated ion channel receptors and G-protein-coupled receptors (GPCR). Fast synaptic transmission is critical for real-time brain functions. Ligand-gated ion channels can handle such rapid processing. These ligand bonding sites can bind to a particular neurotransmitter molecule, open a transmission channel, and activate signaling. There are many G‑protein-coupled receptors, so much so that it comprises one percent of the human genome. These GPCRs form the receptors for neurotransmitters, odorants, lipids, neuropeptides, and large glycoprotein hormones (Davies 2006). Fig. 11 shows the materials and devices that compose the biological nucleotide. 3. Components and devices Components and devices are at the next layer in the full compute stack. At this level, material physics phenomena are captured in a helpful way that can form a building block for a higher-level computing function. On the artificial or electronic side, components consist of various types of transistors, phase-change or memristor memories, optical devices, and switching devices. On the biological side, components include protein complexes that respond to signals, ligand receptors, phosphorylation, and methylation in cytosine-phosphate-guanine (CpG) groups. CpG sites are regions of DNA where a cytosine nucleotide is followed by a guanine nucleotide and can be impacted (silencing genes, switching on or off, or muting them to some degree) by inserting a phosphate group between them. To a degree, one can show a mapping of similar functions between the artificial and biological, but the material and components and devices’ full compute stack layers are implemented with very different materials and physical phenomena to produce the desired effects. At the device level, the basic building block is the neuron. Both artificial and biological neural networks have other support devices included, but the primary focus here is exploring the nature and design of the neuron. A few questions can be asked here to explore a path forward. First, what physical phenomenological properties can be used for the full compute stack components and devices layer functions? The Von Neumann architecture describes the three necessary building blocks for a computing system as (1) a central processing unit (CPU), (2) memory, and (3) input and output devices. Components must exist to enable these functions as they are put together to form a functional and flexible neuron. Second, how are these functions translated into functional parts within a particular design motif, like silicon for artificial neurons and nucleotides in biological neurons? One must consider what materials will the components and devices be built upon. With an extensive legacy of parts and infrastructure, the artificial neuron finds plenty of value in continuing development in silicon. Using this design motif allows a smaller near-term investment. However, in the long term, there will continue to be a big difference between the artificial and biological realization of neurons. a. artificial neuron considerations In the roadmap he developed, Christensen offered many options for the components, devices, and circuits necessary to create an artificial neuron capability, along with functional and efficient neural networks. Synapse transmitters and dendrite receptors interconnections JOHANSEN Human brain function and the creation model 2023 ICC 296
RkJQdWJsaXNoZXIy MTM4ODY=