Uncommon Descent Serving The Intelligent Design Community

Controlling the waves of dynamic, far from equilibrium states: the NF-kB system of transcription regulation.

Share
Facebook
Twitter
LinkedIn
Flipboard
Print
Email

I have recently commented on another thread:

about a paper that (very correctly) describes cells as dynamic, far from equilibrium systems, rather than as “traditional” machines.

That is true. But, of course, the cell implements the same functions as complex machines do, and much more. My simple point is that, to do that, you need much greater functional complexity than you need to realize a conventional machine.

IOWs, dynamic, far from equilibrium systems that can be as successful as a conventional machine, or more, must certainly be incredibly complex and amazing systems, systems that defy everything else that we already know and that we can conceive. They must not only implement their functional purposes, but they must do that by “harnessing” the constantly changing waves of change, of random noise, of improbability. I have commented on those ideas in the mentioned thread, at posts #5 and #8, and I have quoted at posts #11 and #12 a couple of interesting and pertinent papers, introducing the important concept of robustness: the ability to achieve reliable functional results in spite of random noise and disturbing variation.

In this OP, I would like to present in some detail a very interesting system that shows very well what we can understand, at present, of that kind of amazing systems.

The system I will discuss here is an old friend: it is the NF-kB system of transcription factors (nuclear factor kappa-light-chain-enhancer of activated B cells). We are speaking, therefore, of transcription regulation, a very complex topic that I have already discussed in some depth here:

I will remind here briefly that transcription regulation is the very complex process that allows cells to be completely different using the same genomic information: IOWs, each type of cell “reads” differently the genes in the common genome, and that allows the different types of cell differentiation and the different cell responses in the same cell type.

Transcription regulation relies on many different levels of control, that are summarized in the above quoted OP, but a key role is certainly played by Transcription Factors (TFs), proteins that bind DNA and act as activators or inhibitors of transcription at specific sites.

TFs are a fascinating class of proteins. There are a lot of them (1600 – 2000 in humans, almost 10% of all proteins), and they are usually medium sized proteins, about 500 AA long, containing at least one highly conserved domain, the DNA binding domain (DBD), and other, often less understood, functional components.

I quote again here a recent review about human TFs:

The Human Transcription Factors

The NK-kB system is a system of TFs. I have discussed it in some detail in the discussion following the Ubiquitin thread, but I will describe it in a more systematic way here.

In general, I will refer a lot to this very recent paper about it:

Considering Abundance, Affinity, and Binding Site Availability in the NF-kB Target Selection Puzzle

The NF-kB system relies essentially on 5 different TFs (see Fig. 1 A in the paper):

  1. RelA  (551 AAs)
  2. RelB  (579 AAs)
  3. c-Rel  (619 AAs)
  4. p105/p50 (968 AAs)
  5. p100/p52  (900 AAs)

Those 5 TFs work forming dimers, homodimers or heterodimers, for a total of 15 possible compbinations, all of which have been found to work in the cell, even if some of them are much more common.

Then there are at least 4 inhibitor proteins, collectively called IkBs.

The mechanism is apparently simple enough. The dimers are inhibited by IkBs and therefore they remain in the cytoplasm in inactive form.

When an appropriate signal arrives to the cell and is received by a membrane receptor, the inhibitor (the IkB molecule) is phosphorylated and then ubiquinated and detached from the complex. This is done by a protein complex called IKK. The free dimer can then migrate to the nucleus and localize there, where it can act as a TF, binding DNA.

This is the canonical activation pathway, summarized in Fig. 1. There is also a non canonical activation pathway, that we will not discuss for the moment.


Mechanism of NF-κB action. In this figure, the NF-κB heterodimer consisting of Rel and p50 proteins is used as an example. While in an inactivated state, NF-κB is located in the cytosol complexed with the inhibitory protein IκBα. Through the intermediacy of integral membrane receptors, a variety of extracellular signals can activate the enzyme IκB kinase (IKK). IKK, in turn, phosphorylates the IκBα protein, which results in ubiquitination, dissociation of IκBα from NF-κB, and eventual degradation of IκBα by the proteasome. The activated NF-κB is then translocated into the nucleus where it binds to specific sequences of DNA called response elements (RE). The DNA/NF-κB complex then recruits other proteins such as coactivators and RNA polymerase, which transcribe downstream DNA into mRNA. In turn, mRNA is translated into protein, resulting in a change of cell function.

Attribution: Boghog2 at English Wikipedia [Public domain]

Now, the purpose of this OP is to show, in greater detail, how this mechanism, apparently moderately simple, is indeed extremely complex and dynamic. Let’s see.

The stimuli.

First of all, we must understand what are the stimuli that, arriving to the cell membrane, are capable to activate the NF-kB system. IOWs, what are the signals that work as inputs.

The main concept is: the NF-kB system is a central pathway activated by many stimuli:

  1. Inflammation
  2. Stress
  3. Free radicals
  4. Infections
  5. Radiation
  6. Immune stimulation

IOWs, a wide variety of aggressive stimuli can activate the system

The extracellular signal arrives to the cell usually through specific cytokines, for example TNF, IL1, or through pathogen associated molecules, like bacterial lipopolysaccharides (LPS). Of course there are different and specific membrane receptors, in particular IL-1R (for IL1) , TNF-R (for TNF), and many TLRs (Toll like receptors, for pathogen associated structures). A special kind of activation is implemented, in B and T lymphocytes, by the immune activation of the specific receptors for antigen epitopes (B cell receptor, BCR, and T cell receptor, TCR).

The process through which the activated receptor can activate the NF-kB dimer is rather complex: it involves, in the canonical pathway, a macromolecular complex called IKK (IkB kinase) complex, comprising two catalytic kinase subunits (IKKa and IKKb) and a regulatory protein (IKKg/NEMO), and involving in multiple and complex ways the ubiquitin system. The non canonical pathway is a variation of that. Finally, a specific protein complex (CBM complex or CBM signalosome) mediates the transmission from the immune BCR or TCR to the canonical pathway. See Fig. 2:

From: NF-κB Activation in Lymphoid Malignancies: Genetics, Signaling, and Targeted Therapy – Scientific Figure on ResearchGate. Available from: https://www.researchgate.net/figure/Increased-activity-of-the-CARMA1-BCL10-MALT1-signalosome-drives-constitutive-NF-kB_fig2_324089636 [accessed 10 Jul, 2019]
Figure 3 – NF-κB Activation in Lymphoid Malignancies: Genetics, Signaling, and Targeted Therapy
available via license: Creative Commons Attribution 4.0 International

I will not go into further details about this part, but those interested can have a look at this very good paper:

TLR-4, IL-1R and TNF-R signaling to NF-kB: variations on a common theme

In particular, Figg. 1, 2, 3.

In the end, as a result of the activation process, the IkB inhibitor is degraded by the ubiquitin system, and the NK-kB dimer is free to migrate to the nucleus.

An important concept is that this is a “rapid-acting” response system, because the dimers are already present, in inactive form, in the cytoplasm, and must not be synthesized de novo: so the system is ready to respond to the activating signal.

The response.

But what is the cellular response?

Again, there are multiple and complex possible responses.

Essentially, this system is a major regulator of innate and adaptive immune responses. As such, it has a central role in the regulation of inflammation, in immunity, in autoimmune processes, and in cancer.

Moreover, the NF-kB system is rather ubiquitous, and is present and active in many different cell types. And, as we have seen, it can be activated by different stimuli, in different ways.

So, the important point is that the response to activation must be (at least):

  1. Lineage-specific
  2. Stimulus-specific

IOWs, different cells must be able to respond differently, and each cell type must respond differently to different stimuli. That gives a wide range of possible gene expression patterns at the transcription level.

The following paper is a good review of the topic:

Selectivity of the NF-κB Response

For example, IL2 is induced by NF-kB activayion in T cells, but not in B cells (lineage specific response). Moreover, specific cell types can undergo specific, and often different, cell destinies after NF-kB activation: for example, NK-kB is strongly involved in the control and regulation of T and B cell development.

From:

30 years of NF-κB: a blossoming of relevance to human pathobiology

“B and T lymphocytes induce NF-κB in adaptive immune responses through the CARD11:Bcl10:MALT1 (CBM) complex (Hayden and Ghosh, 2008). Newly expressed genes promote lymphocyte proliferation and specific immune functions including antibody production by B cells and the generation of cytokines and other anti-pathogen responses by T cells.”

And, in the same cell type, certain promoters regulated by NF-kB require additional signaling (for example,  in human dendritic cells promoters for Il6Il12b, and MCP-1 require additional p38 histone phosphorylation to be activated), while others can be activated directly (stimulus-specific response).

So, to sum up:

  1. A variety of stimuli can activate the system in different ways
  2. The system itself has its complexity (different dimers)
  3. The response can be widely different, according to the cell type where it happens, and to the type of stimuli that have activated the system, and probably according to other complex variables.
  4. The possible responses include a wide range of regulations of inflammation, of the immune system, of cell specifications or modifications, and so on.

How does it work?

So, what do we know about the working of such a system?

I will ignore, for the moment, the many complexities of the activation pathways, both canonical and non canonical, the role of cyotkines and receptors and IKK complexes, the many facets of NEMO and of the involvement of the ubiquitin system.

For simplicity, we will start with the activated system: the IkB inhibitor has been released from the inactive complex in the cytoplasm, and some form of NF-kB dimer is ready to migrate to the nucleus.

Let’s remember that the purpose of this OP is to show that the system works as a dynamic, far from equilibrium system, rather than as a “traditional” machine. And that such a way to work is an even more amazing example of design and functional complexity.

To do that; I will rely mainly on the recent paper quoted at the beginning:

Considering Abundance, Affinity, and Binding Site Availability in the NF-kB Target Selection Puzzle

The paper is essentially about the NF-kB Target Selection Puzzle. IOWs, it tries to analyze what we know about the specificity of the response. How are specific patterns of transcription achieved after the activation of the system? What mechanisms allow the selection of the right genes to be transcribed (the targets) to implement the specific patterns according to cell type, context, and type of stimuli?

A “traditional” view of the system as a machine would try to establish rather fixed connections. For example, some type of dimer is connected to specific stimuli, and evokes specific gene patterns. Or some other components modulate the effect of NK-kB, generate diversification and specificity of the response.

Well, those ideas are not completely wrong. In a sense, the system does work also that way. Dimer specificity has a role. Other components have a role. In a sense, but only in a sense, the system works as though it were a traditional machine, and uses some of the mechanisms that we find in the concept of a traditional biological machine.

But that is only a tiny part of the real thing.

The real thing is that the system really works as a dynamic, far from equilibrium system, harnessing huge random/stochastic components to achieve robustness and complexity and flexibility of behavior in spite of all those non finalistic parts.

Let’s see how that happens, at least for the limited understanding we have of it. It is important to consider that this is a system that has been studied a lot, for decades, because of its central role in so many physiological and pathological contexts, and so we know many things. But still, our understanding is very limited, as you will see.

So, let’s go back to the paper. I will try to summarize as simply as possible the main concepts. Anyone who is really interested can refer to the paper itself.

Essentially, the paper analyzes three important and different aspects that contribute to the selection of targets at the genomic level by our TFs (IOWs, our NF-kB dimers, ready to migrate to the nucleus. As the title itself summarizes, they are:

  1. Abundance
  2. Affinity
  3. Binding site availability

1. Abundance

Abundance is referred here to two different variables: abundance of NF-kB Binding Sites in the genome and abundance of Nucleus-Localized NF-kB Dimers. Let’s consider them separately.

1a) Abundance of NF-kB Binding Sites in the genome:

It is well known that TFs bind specific sites in the genome. For NF-kB TFs, the following consensus kB site pattern has been found:

 5′-GGGRNWYYCC-3′

where R, W, Y, and N, respectively denote purine, adenine or thymine, pyrimidine, and any nucleotide.

That simply means that any sequence corresponding to that pattern in the genome can, in principle, bind NF-kB dimers.

So the problem is: how many such sequences do exist in the human genome?

Well, a study based on RelA has evaluated about 10^4 consensus sequences in the whole genome, but as NF-kB dimers seem to bind even incomplete consensus sites, the total number of potential binding sites could be nearer to 10^6

1b) Abundance of Nucleus-Localized NF-kB Dimers:

An estimate of the abundance of dimers in the nucleus after activation of the system is that about 1.5 × 10^5 molecules can be found, but again that is derived from studies about RelA only. Moreover, the number of molecules and type of dimer can probably vary much according to cell type.

So, the crucial variable, that is the ratio between binding sites and available dimers, and which could help undertsand the rate of sites saturation in the nucleus, remains rather undecided, and it seems very likely that it can vary a lot in different circumstances.

But there is another very interesting aspect about the concentration of dimers in the nucleus. According to some studies, NF-kB seems to generate oscillations of its nuclear content in some cell types, and those oscillation can be a way to generate specific transcription patterns:

NF-kB oscillations translate into functionally related patterns of gene expression

For example, this very recent paper :

NF-κB Signaling in Macrophages: Dynamics, Crosstalk, and Signal Integration

shows at Fig. 3 the occupancy curve of binding sites at nuclear level after NF-kB activation in two different cell types.

In fibroblasts, the curve is a periodic oscillation, with a frequency that varies according to various factors, and translates into different transcription scenarios accordingly:

Gene expression dynamics scale with the period (g1) and amplitude (g2) of these oscillations, which are influenced by variables such as signal strength, duration, and receptor identity.


In macrophages, instead, the curve is rather:

a single, strong nuclear translocation event which persists for as long as the stimulus remains and tends to remain above baseline for an extended period of time.

In this case, the type of transcription will be probably regulated by the are under the curve, ratehr than by the period and amplitude of the oscialltions, as happened in fibroblasts.

Interestingly, while in previous studies it seemed that the concentration of nuclear dimers could be sufficient to saturate most or all binding sites, that has been found not to be the case in more recent studies. Again from the paper about abundance:

in fact, this lack of saturation of the system is necessary to generate stimulus- and cell-type specific gene expression profiles

Moreover, the binding itself seems to be rather short-lived:

Interestingly, it is now thought that most functional NF-kB interactions with chromatin—interactions that lead to a change in transcription—are fleeting… a subsequent study using FRAP in live cells expressing RelA-GFP showed that most RelA-DNA interactions are actually quite dynamic, with half-lives of a few seconds… Indeed, a recent study used single-molecule tracking of individual Halo-tagged RelA molecules in live cells to show that the majority (∼96%) of RelA undergoes short-lived interactions lasting on average ∼0.5 s, while just ∼4% of RelA molecules form more stable complexes with a lifetime of ∼4 s.

2. Affinity

Affinity of dimers for DNA sequences is not a clear cut matter. From the paper:

Biochemical DNA binding studies of a wide variety of 9–12 base-pair sequences have revealed that different NF-kB dimers bind far more sequences than previously thought, with different dimer species exhibiting specific but overlapping affinities for consensus and non-consensus kB site sequences.

IOWs, we have different dimers (15 different types) binding with varying affinity different DNA sequences (starting from the classical consensus sequence, but including also incomplete sequences). Remember that those sequences are rather short (the consensus sequence is 10 nucleotides long), and that there are thousands of such sequences in the genome.

Moreover, different bindings can affect transcription differently. Again, from the paper:

How might different consensus kB sites modulate the activity of the NF-kB dimers? Structure-function studies have shown that binding to different consensus kB sites can alter the conformation of the bound NF-kB dimers, thus dictating dimer function When an NF-kB dimer interacts with a DNA sequence, side chains of the amino  acids located in the DNA-binding domains of dimers contact the bases exposed in the groove of the DNA. For different consensus kB site sequences different bases are exposed in this groove, and NF-kB seems to alter its conformation to maximize interactions with the DNA and maintain high binding affinity. Changes in conformation may in turn impact NF-kB binding to co-regulators of transcription, whether these are activating or inhibitory, to specify the strength and dynamics of the transcriptional response. These findings again highlight how the huge array of kB binding site sequences must play a key role in modulating the transcription of target genes.

Quite a complex scenario, I would say!

But there is more:

Finally, as an additional layer of dimer and sequence-specific regulation, each of the subunits can be phosphorylated at multiple sites with, depending on the site, effects on nearly every step of NF-kB activation.

IOWs, the 15 dimers we have mentioned can be phosphorylated in many different ways, and that changes their binding affinities and their effects on transcription.

This section of the paper ends with a very interesting statement:

Overall, when considering the various ways in which NF-kB dimer abundances and their affinity for DNA can be modulated, it becomes clear that with these multiple cascading effects, small differences in consensus kB site sequences and small a priori differences in interaction affinities can ultimately have a large impact on the transcriptional response to NF-kB pathway activation.

Emphasis mine.

This is interesting, because in some way it seems to suggest that the whole system acts like a chaotic system, at least at some basic level. IOWs, small initial differences, maybe even random noise, can potentially affect deeply the general working of the whole systems.

Unless, of course, there is some higher, powerful level of control.

3. Availability of high affinity kB binding sequences

We have seen that there is a great abundance and variety of binding sequences for NF-kB dimers in the human genome. But, of course, those sequences are not necessarily available. Different cell types will have a different scenario of binding sites availability.

Why?

Because, as we know, the genome and chromatin are a very dynamic system, that can exist in many different states, continuosly changing in different cell types and, in the same cell type, in different conditions..

We know rather well the many levels of control that affect DNA and chromatin state. In brief, they are essentially:

  1. DNA methylation
  2. Histone modifications (methylation, acetylation, etc)
  3. Chromatin modifications
  4. Higher levels of organization, including nuclear localization and TADs (Topologically Associating Domains)

For example, from the paper:

The promoter regions of early response genes have abundant histone acetylation or trimethylation prior to stimulation [e.g., H3K27ac, (67) and H4K20me3, (66)], a chromatin state “poised” for immediate activation…  In contrast, promoters of late genes often have hypo-acetylated histones, requiring conformational changes to the chromatin to become accessible. They are therefore unable to recruit NF-kB for up to several hours after stimulation (68), due to the slow process of chromatin remodeling.

We must remember that each wave of NK-kB activation translates into the modified transcription of a lot of different genes at the genome level. It is therefore extremely important to consider what genes are available (IOWs, their promoters can be reached by the NF-kB signal) in each cell type and cell state.

The paper concludes:

Taken together, chromatin state and chromatin organization strongly influence the selection of DNA binding sites by NF-kB dimers and, most likely, the selection of the target genes that are regulated by these protein-DNA interaction events. Analyses that consider binding events in the context of three-dimensional nuclear organization and chromatin composition will be required to generate a more accurate view of the ways in which NF-kBDNA binding affects gene transcription.

This is the main scenario. But there are other components, that I have not considered in detail for the sake of brevity, for example competition between NF-kB dimers and the complex role and intervention of other co-regulators of transcription.

Does the system work?

But does the system work?

Of course it does. It is a central regulator, as we have said, of many extremely important biological processes, above all immunity. This is the system that decides how immune cells, T and B lymphocytes, have to behave, in terms of cell destiny and cell state. It is of huge relevance in all inflammatory responses, and in our defense against infections. It works, it works very well.

And what happens if it does not work properly?

Of course, like all very complex systems, errors can happen. Those interested can have a look at this recent paper:

30 years of NF-κB: a blossoming of relevance to human pathobiology

First of all, many serious genetic diseases have been linked to mutations in genes involved in the system. You can find a list in Table 1 of the above paper. Among them, for example, some forms of SCID, Severe combined immunodeficiency, one of the most severe genetic diseases of the immune system.

But, of course, a dysfunction of the NF-kB system has a very important role also in autoimmune diseases and in cancer.

Conclusions.

So, let’s try to sum up what we have seen here in the light of the original statement about biological systems that “are not machines”.

The NF-kB system is a perfect example. Even if we still understand very little of how it works, it is rather obvious that it is not a traditional machine.

A traditional machine would work differently. The signal would be transmitted from the membrane to the nucleus in the simplest possible way, without ambiguities and diversions. The Transcription Factor, once activated, would bind, at the level of the genome, very specific sites, each of them corresponding to a definite cascade of specific genes. The result would be clear cut, almost mechanical. Like a watch.

But that’s not the way things happen. There are myriads of variations, of ambiguities, of stochastic components.

The signal arrives to the membrane in multiple ways, very different one from the other: IL1, IL17, TNF, bacterial LPS, and immune activation of the B cell receptor (BCR) or the T cell receptor (TCR) are all possible signals.

The signal is translated to the NF-kB proteins in very different ways: canonical or non canonical activation, involving complex protein structures such as:

The CBM signalosome, intermediate between immune activation of BCR or TCR and canonical activation of the NF-kB. This complex is made of at least three proteins, CARD11, Bcl10 and MALT1.

The IKK complex in canonical activation: this is made of three proteins, IKK alpha, IKK beta, and NEMO. Its purpose is to phosphorylate the IkB, the inhibitor of the dimers, so that it can be ubiquinated and released from the dimer. Then the dimer can relocate to the nucleus.

Non canonical pathway: it involves the following phosphorylation cascade: NIK -> IKK alpha dimer -> Relb – p100 dimer -> Relb – p50 dimer (the final TF). It operates during the development of lymphoid organs and is responsible for the generation of B and T lymphocytes.

Different kinds of activated dimers relocate to the nucleus.

Different dimers, in varying abundance, interact with many different binding sites: complete or incomplete consensus sites, and probably others. The interaction is usually brief, and it can generate an oscillating pattern, or a more stable pattern

Completely different sets of genes are transcribed in different cell types and in different contexts, because of the interaction of NF-kB TFs with their promoters.

Many other factors and systems contribute to the final result.

The chromatin state of the cell at the moment of the NF-kB activation is essential to determine the accessibility of different binding sites, and therefore the final transcription pattern.

All these events and interactions are quick, unstable, far from equilibrium. A lot of possible random noise is involved.

In spite of that amazing complexity and potential stochastic nature of the system, reliable transcripion regulation and results are obtained in most cases. Those results are essential to immune cell differentiation, immune response, both innate and adaptive, inflammation, apoptosis, and many other crucial cellular processes.

So, let’s go back to our initial question.

Is this the working of a machine?

Of course it is! Because the results are purposeful, reasonably robust and reliable, and govern a lot of complex processes with remarkable elegance and efficiency.

But certainly, it is not a traditional machine. It is a lot more complex. It is a lot more beautiful and flexible.

It works with biological realities and not with transistors and switches. And biological realities are, by definition, far from equilibrium states, improbable forms of order that must continuously recreate themselves, fighting against the thermodynamic disorder and the intrinsic random noise that should apparently dominate any such scenario.

It is more similar to a set of extremely clever surfers who succeed in performing elegant and functional figures and motions in spite of the huge contrasting waves.

It is, from all points of view, amazing.

Now, Paley was absolutely right. No traditional machine, like a watch, could ever originate without design.

And if that is true of a watch, with its rather simple and fixed mechanisms, how much truer it must be for a system like NF-kB? Or, for that, like any cellular complex system?

Do you still have any doubts?

Added graphic: The evolutionary history, in terms of human conserved information, of the three proteins in the CBM signalosome.
On the y axis, homologies with the human protein as bits per aminoacid (bpa). On the x axis, approximate time of appearance in million of years.
The graphic shows the big information jump in vertebrates for all three protens , especially CARD11.


Added graphic: two very different proteins and their functional history


Added graphic (for Bill Cole). Functional history of Prp8, collagen, p53.
Comments
NF-kB Signaling Pathways in Osteoarthritic Cartilage Destruction https://www.mdpi.com/2073-4409/8/7/734OLV
October 9, 2019
October
10
Oct
9
09
2019
02:16 AM
2
02
16
AM
PDT
NF-kB Signaling in Ovarian Cancer https://www.mdpi.com/2072-6694/11/8/1182OLV
October 9, 2019
October
10
Oct
9
09
2019
02:11 AM
2
02
11
AM
PDT
Single-Cell Analysis of Multiple Steps of Dynamic NF-kB Regulation in Interleukin-1a-Triggered Tumor Cells Using Proximity Ligation Assays https://www.mdpi.com/2072-6694/11/8/1199OLV
October 9, 2019
October
10
Oct
9
09
2019
02:04 AM
2
02
04
AM
PDT
Adenovirus early region 3 RIDa protein limits NFkB signaling through stress-activated EGF receptors PLoS Pathog. 2019 Aug; 15(8): e1008017. Published online 2019 Aug 19. doi: 10.1371/journal.ppat.1008017 EGFRs activated by stress of adenoviral infection regulated signaling by the NFkB family of transcription factors, the NFkB p65 subunit was phosphorylated at Thr254 RIDa expression was sufficient to down-regulate the same EGFR/NFkB signaling axis https://journals.plos.org/plospathogens/article?id=10.1371/journal.ppat.1008017OLV
October 9, 2019
October
10
Oct
9
09
2019
01:58 AM
1
01
58
AM
PDT
DATCG: Hi, nice to hear from you! :) Good arguments, as usual. And interesting links. By the way, IDPs seem to have a special role in innate immunity, as a tool against viruses and bacteria, as shown by the following papers: Intrinsic disorder in proteins involved in the innate antiviral immunity: another flexible side of a molecular arms race. https://www.ncbi.nlm.nih.gov/pubmed/24184279
Abstract We present a comprehensive bioinformatics analysis of the abundance and roles of intrinsic disorder in human proteins involved in the antiviral innate immune response. The commonness of intrinsic disorder and disorder-based binding sites is evaluated in 840 human antiviral proteins and proteins associated with innate immune response and defense response to virus. Among the mechanisms engaged in the innate immunity to viral infection are three receptor-based pathways activated by the specific recognition of various virus-associated patterns by several retinoic acid-inducible gene I-like receptors, toll-like receptors, and nucleotide oligomerization domain-like receptors. These modules are tightly regulated and intimately interconnected being jointly controlled via a complex set of protein-protein interactions. Focused analysis of the major players involved in these three pathways is performed to illustrate the roles of protein intrinsic disorder in controlling and regulating the innate antiviral immunity. We mapped the disorder into an integrated network of receptor-based pathways of human innate immunity to virus infection and demonstrate that proteins involved in regulation and execution of these innate immunity pathways possess substantial amount of intrinsic disorder. Disordered regions are engaged in a number of crucial functions, such as protein-protein interactions and interactions with other partners including nucleic acids and other ligands, and are enriched in posttranslational modification sites. Therefore, host cells use numerous advantages of intrinsically disordered proteins and regions to fight flexible invaders and viruses and to successfully overcome the viral invasion. And: Abundance and functional roles of intrinsic disorder in the antimicrobial peptides of the NK-lysin family https://www.tandfonline.com/doi/abs/10.1080/07391102.2016.1164077
Abstract NK-lysins are antimicrobial peptides (AMPs) that participate in the innate immune response and also have several pivotal roles in various biological processes. Such multifunctionality is commonly found among intrinsically disordered proteins. However, NK-lysins have never been systematically analyzed for intrinsic disorder. To fill this gap, the amino acid sequences of NK-lysins from various species were collected from UniProt and used for the comprehensive computational analysis to evaluate the propensity of these proteins for intrinsic disorder and to investigate the potential roles of disordered regions in NK-lysin functions. We analyzed abundance and peculiarities of intrinsic disorder distribution in all-known NK-lysins and showed that many NK-lysins are expected to have substantial levels of intrinsic disorder. Curiously, high level of intrinsic disorder was also found even in two proteins with known 3D-strucutres (NK-lysin from pig and human granulysin). Many of the identified disordered regions can be involved in protein–protein interactions. In fact, NK-lysins are shown to contain three to eight molecular recognition features; i.e. short structure-prone segments which are located within the long disordered regions and have a potential to undergo a disorder-to-order transition upon binding to a partner. Furthermore, these disordered regions are expected to have several sites of various posttranslational modifications. Our study shows that NK-lysins, which are AMPs with a set of prominent roles in the innate immune response, are expected to abundantly possess intrinsically disordered regions that might be related to multifunctionality of these proteins in the signal transduction pathways controlling the host response to pathogenic agents.
gpuccio
October 5, 2019
October
10
Oct
5
05
2019
04:59 AM
4
04
59
AM
PDT
And to follow-up... on IDPs, Why the word, "disorder?" Press Release... Supposed disorder NOT disorder after all Paper: Phosphorylation orchestrates the structural ensemble of the intrinsically disordered protein HMGA1a and modulates its DNA binding to the NF-kB promoter DATCG
October 4, 2019
October
10
Oct
4
04
2019
09:32 PM
9
09
32
PM
PDT
Why use the word, "intrinsically" to describe a Flexible Protein? Because, they must remain ever vigilant to force Design out and keep blind, unguided mutations in.
intrinsically [in?trinz?k(?)l?, in?trins?k(?)l?] ADVERB - in an essential or natural way.
As if these proteins are not designed to be flexible and other proteins to be rigid, ordered and limited match by design. By being close-minded to only a natural cause, they become blind to Design.DATCG
October 4, 2019
October
10
Oct
4
04
2019
05:22 PM
5
05
22
PM
PDT
Gpuccio, There's new article out about Intrinsically Disordered Proteins :) cited by Evolution News, woot, woot! So, thought I'd look IDPs up for Immune System which you've said you will be posting on in the future. Because certainly the Immune System needs Flexible Proteins, right? Here's an article on first search:
"The scientists, nine of whom are from Johns Hopkins and one from the University of Houston, set out to answer that question. They chose for their study a disordered protein taken from human cells called glucocorticoid receptor, which regulates genes that control, among other functions, metabolism and immune system response.
https://elifesciences.org/articles/30688 Smiles... IDPs - misnamed are Flexible Proteins, designed to be flexible. :) It goes on, sorry this is off-topic I know, but fun! We live in amazing times of incredible Design Discovery :)
By manipulating segments of the protein in the lab, they were able to show how one portion acts on another, and that the disordered protein creates versions of itself to act almost in place of regulator molecules that govern its activity. The disordered protein uses an activation-repression dynamic—described by Hilser as similar to attracting and repelling magnets—between sections within the disordered chain to regulate its own activities and those of other proteins. "Our work uncovered the language of how these spaghetti pieces communicate," Hilser said. "We showed that those pieces of spaghetti interact with each other sort of like attracting and repulsing magnets, creating a kind of tug-of-war, and that the body can make different versions of the protein to tune which part wins the tug-of-war." Yet to be explained, he said, is how the interactions among these proteins and the sub-sections happen and how all this can ultimately be used to treat disorders that emerge when things go awry with these molecules central to almost all life function.
Looking forward to your post on the Immune System and FI!DATCG
October 4, 2019
October
10
Oct
4
04
2019
05:11 PM
5
05
11
PM
PDT
Jawa: Yes, it's been a long time! :) Now I am really working hard at the OP about the adaptive immune system. Indeed, I think it will have to be in two separate parts. The first part will be about the generation of the basic antibody repertoire, let's say the pre-antigen part. The second part will be about affinity maturation and other related post-antigen topics. The adaptive immune system is really an amazing example of protein engineering lab. That anybody in the world really believe that it is not an engineered system is completely beyond my understanding. Moreover, the engineering principles that govern the two different aspects I have mentioned are equally complex and efficient, but completely different in their working and perspective. Really fascinating stuff. And, luckily, much more is understood today than even a few years ago.gpuccio
October 4, 2019
October
10
Oct
4
04
2019
08:30 AM
8
08
30
AM
PDT
Is this discussion still in the list? Popular Posts (Last 30 Days)
Controlling the waves of dynamic, far from… (1,290) Does The Bible “condone” slavery, even… (1,220) Sean Carroll: “Nowadays, when a more… (1,067) A world-famous chemist tells the truth:… (968) Researcher: Evidence for early man in Asia half a… (964)
It's almost 3 months old. It has been visited 6,329 times.jawa
October 4, 2019
October
10
Oct
4
04
2019
12:20 AM
12
12
20
AM
PDT
GP
Of course we get a signal even after 60 – 100 million years only of separation (humans vs mice or other mammals). However, it is mixed with a non trivial component of passive conservation. It is a valid signal, however, and the different behaviour of different proteins can be observed just the same. I prefer to discuss the vertebrate transition because in that case I am more confident that most, or practically all of the conservation can be attributed to fnctional constraint.
I agree. These bottlenecks where additional functional constraint occur is very interesting.bill cole
September 26, 2019
September
09
Sep
26
26
2019
03:41 PM
3
03
41
PM
PDT
Bill: Of course we get a signal even after 60 - 100 million years only of separation (humans vs mice or other mammals). However, it is mixed with a non trivial component of passive conservation. It is a valid signal, however, and the different behaviour of different proteins can be observed just the same. I prefer to discuss the vertebrate transition because in that case I am more confident that most, or practically all of the conservation can be attributed to fnctional constraint. Prp8 is really an amazing example of extreme conservation. TFs like p53 are probably not the best way to assess neutral divergence, because we know quite well that they are usually bimodal: highly conserved functional DBDs and poorly conserved functional sequences implied in complex interactions and regulations. A more "neutral" reference could be some structural protein, like collagen, even if those protein too have probably specific roles in different species. For your convenience, I am adding at the end of the OP another graph, with the evolutionary history, always in terms of human conserved sequence similarity, of these three proteins: Prp8, p53 and collagen. As you can see, both p53 and collagen have a rather "neutral" behaviour, even if probably for different reasons, as discussed. Both behave, in average, as the mean of the whole proteome. However, some minor adjustment seems to take place for both at the vertebrate transition (but that is true, in general, for the whole proteome too), and collagen seems to show some jump after marsupials. But, in general, the behaviour of collagen can be probably attributed mainly to neutral divergence, and for p53 there is probably the effect of a dual, inverse, functional effect. The really amazing thing is Prp8, which, as we know, exhibits practically the same sequence starting at the beginning of metazoa. The cnidarian protein has indeed 92.23% identity with the human protein. Which is then conserved, and only slightly refined, in all other metazoa. Amazing, especially for a protein which is 2335 AA long!gpuccio
September 26, 2019
September
09
Sep
26
26
2019
04:14 AM
4
04
14
AM
PDT
GP
For my purposes, alignments bewteen humans and primates, or even other mammals like mice, are not really interesting, because for those species the split is toot recent. If I get 95% similarity between humans and mice, for example, I really cannot say how much of that is functional constraint and how much is passive similarity. Very simply, there has not been enough time for neutral varitaion to act significantly where it can act.
What would you expect from neutral mutations over 60 million years like the split between humans and mice? We get 99.9% identity between humans and mice with prp8. We only get 89% identity with the TTN protein. P53 we get 76% identical positions. This looks to me like different designs or more neutral mutations tolerated. Will look at it more tomorrow. Trying to think about how many neutral mutations we would expect if the protein could tolerate mutations in every location. No need to respond at this point just thinking out loud.bill cole
September 25, 2019
September
09
Sep
25
25
2019
04:46 PM
4
04
46
PM
PDT
Bill at #675: Please, read also my #676. For my purposes, alignments bewteen humans and primates, or even other mammals like mice, are not really interesting, because for those species the split is toot recent. If I get 95% similarity between humans and mice, for example, I really cannot say how much of that is functional constraint and how much is passive similarity. Very simply, there has not been enough time for neutral varitaion to act significantly where it can act. That's why I look at least at cartilaginous fish (400+ million years of separations). Of course single celled eukaryotes are very distant, and probably different functional constraints can have a big role there, especially for regulatory proteins. Another problem is that many proteins exist in rather different isoforms, with rather different functions. Let's consider, for example, Prickle 1 and Prickle 2, which I have discussed in an older OP. They are 800+ AA long, but they share, in humans, "only" 711 bits. They are similar, but they are different proteins, with different roles, different tissue specificity, and so on. For comparison, Prickle 1 in humans and Prickle 1 in cartilaginous fishes share 1189 bits of sequence similarity, after 400+ million years of separation. This is a way of detecting how a difference between two similar proteins has certainly, in some cases, a heavy functional meaning.gpuccio
September 25, 2019
September
09
Sep
25
25
2019
02:46 PM
2
02
46
PM
PDT
Bill: You say: "We also need to take into account that you are sampling many different applications of the protein across many different complexity of species." Correct. But that is one of the reasons that my measures are underestimating FI, not the opposite. When a specific sequence is conserved through hundreds of millions of years, it's beacuse it is highly functional, and it is needed in practically all the different adaptations of that protein in different species. The parts of the sequence which are not conserved, of course, are a mix of neutral or quasi neutral components, which are relatively free to change, and of other functional, maybe even highly functional, components, which need to change to adapt the protein to different contexts. It is not difficult to conceive those changing functional components. Proteins have many "signals" that are context specific: localization signals, parallel interactions with regulatory networks, phosphorylation sites, ubiquitination sites, and so on. Many of those signals are different in different species, even if the general function of the protein may appear similar. Some proteins are just machines that perform more or less the same thing in different species. But most proteins are not like that. Especially proteins that are involved in complex networks. Like most proteins that I have discussed here. Like most proteins that are long and complex, and which present clear information jumps at specific evolutionary nodes, like the transition to vertebrates. That's why species are different, and each of them is infinitely complex, but in different ways. So, it's not really a "sampling" that is at work in my procedure. I measure what is conserved thorugh long evolutionary times. That is not really a sampling, but a direct measure. You see, the simple truth is that here the background noise is not so important. Big samples are useful to detect a signal against the background noise. But when we observe signals that have the size of hundreds of bits of sequence similarities, the background noise of random similarities is completely irrelevant. Do you know why I uise the bitscore, and not the E value, which is another output os Blast? The E value is in itself a measure of probability, so it would be very simple to use directly that value. From the Blast FAQ: "The Expect value (E) is a parameter that describes the number of hits one can "expect" to see by chance when searching a database of a particular size. It decreases exponentially as the Score (S) of the match increases. Essentially, the E value describes the random background noise. " well, the simple reason that I cannot use the E value instead of the bitscore is that, for any of the protein comparisons that I discuss, which usually have bitscores of hundreds of bits, the E value is simply rounded to zero. For example: ATP synthase beta chain between E. coli and humans: bitscore = 663 bits. E value = 0 RAG1 between cartilaginous fishes and humans: bitscore = 1361 bits. E value = 0 So, as you can see, the bitscore can distinguish between proteins of different conserved complexity, while the E value (which measures the background noise) is flattened to 0 by the algorithm itself. Because the background noise is irrelevant, at those levels of signal.gpuccio
September 25, 2019
September
09
Sep
25
25
2019
02:31 PM
2
02
31
PM
PDT
Hi GP Great explanation and reviewing the points. They have gone from you cannot make the measurement to too small of sampling as I pointed out this statistical technique used in polling. Now that population size does not matter if we take 9 samples than the error is 33%. 1/the square root of 9. By my arithmetic the error is less than one bit in each direction. Do you agree? I did some alignment work this am and found some interesting results. Different types of yeast have 70% alignment to prp8. If we look at humans and monkeys the alignment is 100% if we include mice it is 99.9%. When we compare yeast to slime mold we get 56 to 59% alignment same for mammals vs yeast. The issue is the functional constraint in yeast is much lower than in multicellular organisms. The functional constraint in earlier multicellular eukaryotic cells appears to be less than in mammals. This looks like design changes to me. What do you think?bill cole
September 25, 2019
September
09
Sep
25
25
2019
02:23 PM
2
02
23
PM
PDT
Bill: I am not sure I understand your point about sampling. In general, it is true that the accuracy of an estimate depends on the sample size and not on the population size, provided that the population is big enough. But I am not sure of what this has to do with the context of our discussions. The problem in evaluating FI is usually to have some good estimate of the target sice. The search size can be usually estimated easily enough. And then same can be said of the probabilistic resources of the system. Remember, we don't need high precision here. We are dealing with very big numbers, and what really matters is the approximate order of magnitude of the target space/search space ratio. In general, functional spaces are extremely small if compared to search spaces, especially as the search space grows exponentially. That is rather evident in language and in software, for example. Instead, neo-darwinists (OK, I have decided that I will go on calling them that way :) ) try desperately to convince us (and probably themselves) that the functional space of proteins os special, that complex functions are connected (as if a watch were connected to a rifle, or to a book) and that complex functions abund so much in that space that they will be found like mushrooms. Or, in alternative, that complex functions are really simple when they first appear, as if a watch could evolve from one gross quasi-gear, a book from some random sign made by the wind, and so on. Of course, this is all mere imagination and dogma. Things are not that way. Complex functions are complex exactly because they require a lot of specific bit configurations. And complex functions do not emerge from simple functions. They emerge out of planning, out of understanding, out of purpose. The 2000 protein superfamilies are not a connected space where it is easy to go from one function to the other. Not at all. It is not a case that such transitions are never observed. They are never observed because they do not exist. These things are so obvious that every thinking person would admit them, if they were not, at the same time, too dangerous for the current ideology. So, we witness the sad show of intelligent people sticking to impossible explanations, and to the obstinate denial of very simple and powerful concepts like FI, and so on. So we see depressing "discussions" to deny that FI exists, or to affirm that it is everywhere, or to build it by summing simple bits of simple functions. How sad. If FI is such a wrong idea, why was Szostak so interested in it? Nobody knows. If function is so abundant in the protein functional space, why did the same Szostak have to build a big random library just to find some weak and useless ATP binding, and then engineer it thorugh random variation and intelligent selection just to get some stromg and equally useless ATP binding? If function is so abundant, why was it impossible for the researchers of the rugged landscape paper to find the functional island of the wildtype sequence? And so on, and so on. All the scenario of protein engineering shows how difficult it is to manifacture a functional protein, even using all our knowledge of biochemical laws and of the protein landscape, and all our most recent technology, and of course all possible imitations of what already exists in living beings. Yes, because, even using design, the best design we are able to produce, it is really difficult to engineer proteins. But certainly it must be us; we must be really dumb, given that complex functional proteins grow in the search space like mushrooms.gpuccio
September 25, 2019
September
09
Sep
25
25
2019
02:01 PM
2
02
01
PM
PDT
Apparently T aquaticus thinks that science can be reduced to a parlor game.
Does this sequence have FI, and was it made by a mind? GVGICQSWMFVQKKMDCIGLCIPMIIMMIQGSSAYTKHKMAFTPRNSNLAFMVHHISQWG SGDARVDAEMQINKPQWLNEKNGNTHFNEYFMGDMYDQIGRKTRNQSGDFSGFALPCFFY TEYRNCHRLRIGNHRRNYFTHKYCSKEWPVFPCGPYFSKNDFGIMSYHQYSTALSHECLV TAGEHDHFQSNIKIMMHEYS
CONTEXT is important here. Clearly if we observed those letters scratched into a cave wall we would infer it was the product of a mind. That is an intelligent agent did it. If we received that on a radio channel we would infer there was a mind behind it. Context matters in science. Hopefully PS has something better to offer than sheer desperation. So far they do not. With respect to biology the relevant sequences produce observed functionality. The point is to determine whether said functionality arose by blind and mindless processes or by intelligent design. gpuccio, and many others, say that by measuring the information contained in that functional sequence such a determination can be made. These clowns think we should be able to predict that when handed random sequences, which sequences will be functional. When all that is happening is that we can predict which sequences required intelligent design.ET
September 25, 2019
September
09
Sep
25
25
2019
10:27 AM
10
10
27
AM
PDT
Gpuccio Thanks for the response. I completely agree with you. Something came up this morning in a discussion with Rum. His claim is that the possible sequences are too large for and estimate. When I looked at political poll sampling strategy it was interesting that by their methods accuracy is dependent on sample size and not the size of the population. I would like to understand this better as this is very interesting for your method. We also need to take into account that you are sampling many different applications of the protein across many different complexity of species.bill cole
September 25, 2019
September
09
Sep
25
25
2019
10:02 AM
10
10
02
AM
PDT
Bill: In the past, I have already made an explicit challenge (to those at TSZ, I believe) to offer any sequence so that I could infer, or not infer, design for it, without any false positive. Friends here at UD were ready to offer many examples of functional sequences for which I readily and correctly inferred design. Interlocutors from the other field tried all sorts of tricks, more or less on the line of T, trying to show, I don't know for what reason, that I could not detect design in all possible cases, or that there were many strings for which I could simply not infer design, without knowing if they were true negatives or false negatives. Which are of course very trivial truths, that I could have agreed upon in advence. Design inference is about inferring with extremely high certainty that some specific object is designed. It is not about recognizing all designed objects. It is not about recognizing all non designed objects. It is a procedure with virtually no false positive (if the threshold of FI is chosen appropriately), and with many, many false negatives. Again, these are really the basics of ID theory.gpuccio
September 25, 2019
September
09
Sep
25
25
2019
02:42 AM
2
02
42
AM
PDT
Bill: This is all we can say, IMO: 1) It is a sequence of 200 characters, with no apparent linguistic meaning. 2) It can easily be intepreted as a sequence of 200 AAs (the characters are those of the AA one letter code). 3) I have blasted it, without finding any significant similarity. That said, the sequence appears to implement no specific function of any complexity, as far as I can see. So I agree with you, it has apparently extremely low FI (we could always define some very generic function for a non specific 200 character sequence). Therefore, there is no indication at all to infer design for the sequence. It is a negative, with the information we have about it. Now, there are of course two possibilities: a) It is a true negative. b) It is a false negative. There is some function for it, either as a character sequence or as a sequence of AAs, but we have not recognized it. In both cases, there is no problem. As should be very clear to whoever has some basic understanding of ID and of the design inference, the procedure for design inference is conceived to have practically no false positives, but many false negatives. Examples of false negatives are: a) Designed objects whose function is very simple, and does not qualify for a safe design inference. In a general sense, anything below 500 bits of FI, or any other appropriate threshold for the system that is being considered. b) Designed objects whose function is not recognized by the examiner. Really, it is rather tiresome to have to explain all the basics to people who arrogantly believe that they have already understood all, while they don't even know what they are speaking of.gpuccio
September 25, 2019
September
09
Sep
25
25
2019
02:25 AM
2
02
25
AM
PDT
Gpuccio and all. Here is a post from PS I found interesting. Any thoughts? My thoughts are 0 FI unless the function can be clearly defined. Shows that minds are capable of FI and garbage :-) T aquaticus Friendly Atheist Biologist 6m colewd: The functional information inside DNA to start. Ok. Does this sequence have FI, and was it made by a mind? GVGICQSWMFVQKKMDCIGLCIPMIIMMIQGSSAYTKHKMAFTPRNSNLAFMVHHISQWG SGDARVDAEMQINKPQWLNEKNGNTHFNEYFMGDMYDQIGRKTRNQSGDFSGFALPCFFY TEYRNCHRLRIGNHRRNYFTHKYCSKEWPVFPCGPYFSKNDFGIMSYHQYSTALSHECLV TAGEHDHFQSNIKIMMHEYS How did you calculate FI, and how does the calculation test your hypothesis?bill cole
September 24, 2019
September
09
Sep
24
24
2019
11:37 AM
11
11
37
AM
PDT
Communication codes in developmental signaling pathways. Review article Li P, et al. Development. 2019. https://dev.biologists.org/content/146/12/dev170977.long
Abstract A handful of core intercellular signaling pathways play pivotal roles in a broad variety of developmental processes. It has remained puzzling how so few pathways can provide the precision and specificity of cell-cell communication required for multicellular development. Solving this requires us to quantitatively understand how developmentally relevant signaling information is actively sensed, transformed and spatially distributed by signaling pathways. Recently, single cell analysis and cell-based reconstitution, among other approaches, have begun to reveal the 'communication codes' through which information is represented in the identities, concentrations, combinations and dynamics of extracellular ligands. They have also revealed how signaling pathways decipher these features and control the spatial distribution of signaling in multicellular contexts. Here, we review recent work reporting the discovery and analysis of communication codes and discuss their implications for diverse developmental processes.
OLV
September 22, 2019
September
09
Sep
22
22
2019
06:39 PM
6
06
39
PM
PDT
Hugh Kenneth @664:
UD is a news aggregator site where someone is payed to post links to several articles every day.
Are you sure that that statement is exactly valid? :) The comparison between those websites was done basically motivated by curiosity about the generated traffic according to Alexa data. And it was done mainly is relation to GP having a discussion with folks who contribute to the other mentioned websites. Very simple. However, since you brought up the “news aggregator” parameter, should UD be compared with DR for example? Yes, why not? :) I literally predicted that PS could have a noticeable increase in traffic after GP was commenting in their site. Well, the jump was more impress than I expected. Having a discussion with GP was definitely a boost to PS and PT. The timing of the numbers shown in Alexa confirms this. :) Do you agree? BTW, here’s another way to look at the traffic issue: Popular Posts (Last 30 Days) Rare hominin skull upsets tidy human origins theory (3,669) Controlling the waves of dynamic, far from… (2,920) Once More from the Top on “Mechanism” (1,802) Chemist James Tour calls time out on implausible… (1,337) Does The Bible “condone” slavery, even… (1,205) Are those 5 OPs comparable as apples?jawa
September 22, 2019
September
09
Sep
22
22
2019
10:45 AM
10
10
45
AM
PDT
Hugh Kenneth @664:
But even comparing the sites you did is not exactly comparing apples and apples either. UD is a news aggregator site where someone is payed to post links to several articles every day.
Do you mean “news aggregator” like “Drudgereport.com”? :) For example, couldn’t the following websites be comparable as news media? Alexa ranking: wsj.com 472 Drudgereport.com 639 Economist.com 3,183 I agree that there are several interesting parameters shown in the information provided by Alexa. Perhaps you understand them better than I do. Having GP explaining some ID concepts to the guys at PS definitely provides a tempting excuse to somehow compare UD with PS. They seem more closer in terms of covered topics than DR and UD are, even though both are in the category of news aggregators, according to your commentary. Seeing in PS folks who seem active contributors at PT and TSZ definitely creates another tempting excuse to put PT, TSW, PS in the same group. Seeing all those folks engaged in public discussion with GP adds support to such argument of “comparable”.jawa
September 22, 2019
September
09
Sep
22
22
2019
10:16 AM
10
10
16
AM
PDT
GP
Sequence comparisons allow us, in the right context, to detect FI even when the function cannot be implemented, like in the case of pseudogenes.
Thanks. This really helped me solidify the concept in my mind. Even though Rum and T did not have the right concept this exercise of explaining it to them really helped. I think they are on the same page with us at this point. We will see. I reviewed (re read the conversation this AM) your discussion with Josh Art and Steve. I think if we do it again we should put the 500 bit discussion to the side as we can conclude whether it is design or evolutionary processes latter once we get agreement that the FI measurement is reasonable. Since Kirk is working to refine his methods right now it would be interesting if we could test correlation and repeatability between your results and his. Just a thought.bill cole
September 21, 2019
September
09
Sep
21
21
2019
03:00 PM
3
03
00
PM
PDT
Jawa@659, yes, I do understand your point, and I admit using an absurd example to emphasize my point. But even comparing the sites you did is not exactly comparing apples and apples either. UD is a news aggregator site where someone is payed to post links to several articles every day. From what I have seen, these generally don’t trigger a lot of comments, but they do often lead to related OPs that do. To the best of my knowledge, these other sites limit themselves to relatively infrequent OPs. I would also be interested in what the traffic numbers mean. Are they simply hits, are they unique hits, are they filtered by duration on site? The reason I ask is two-fold. First, I compared the bounce rate of UD against a couple of the others on your list and UD’s bounce rate is much higher, suggesting lower average meaningful engagement by those that visit the sight. Second, there are a relatively few unique commenters who post comments many times every day. People like BA77, KF, Gpucio and ET. They alone are probably responsible for a not inconsequential fraction of UD traffic.Hugh Kenneth
September 21, 2019
September
09
Sep
21
21
2019
07:48 AM
7
07
48
AM
PDT
Bill: We absolutely agree on the concept. My way of expressing it is that FI is a property of the function, not of the object. The object may implement the function or not. For the function to be implemented, by definition, all the necessary bits must be there (the complete FI). If an object has almost all the necessary bits, it has most of the FI linked to the function, but it cannot yet implement the function. So, I would not say that the object has zero FI, but that it has almost all, but not all, the FI linked to the function, but cannot yet implement the function. The substance is the same, but this is the terminology I prefer to use. Sequence comparisons allow us, in the right context, to detect FI even when the function cannot be implemented, like in the case of pseudogenes. Really, all this discussion generated by Rumracket and company is completely meaningless, and an obvious and desperate attempt to deny the evident truth.gpuccio
September 21, 2019
September
09
Sep
21
21
2019
04:40 AM
4
04
40
AM
PDT
GP, How are the epigenetic markers associated with determining which DNA part can be expressed within a particular cell type? what’s the relation or association between the TFs and the epigenetic markers and the ncRNAs that serve as regulatory elements ? How’s that choreography composed before it’s needed? Thanks.OLV
September 21, 2019
September
09
Sep
21
21
2019
02:22 AM
2
02
22
AM
PDT
Does anybody know a plant biologist who could explain this to us here? Who could answer some questions we may have about this paper? Single-cell three-dimensional genome structures of rice gametes and unicellular zygotes https://www.nature.com/articles/s41477-019-0471-3
Our results reveal specific 3D genome features of plant gametes and the unicellular zygote, and provide a spatial chromatin basis for ZGA and epigenetic regulation in plants.
OLV
September 21, 2019
September
09
Sep
21
21
2019
02:13 AM
2
02
13
AM
PDT
1 2 3 4 5 25

Leave a Reply