1 / 32

Natural and “Artificial” Immune Systems

Natural and “Artificial” Immune Systems. This is lecture 19 of Biologically Inspired Computing; about Natural and Artificial Immune Systems. It borrows much from a tutorial presentation by Jon Timmis, now at York. Overview. What are Artificial Immune Systems? Background immunology

studs
Télécharger la présentation

Natural and “Artificial” Immune Systems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Natural and “Artificial” Immune Systems This is lecture 19 of Biologically Inspired Computing; about Natural and Artificial Immune Systems. It borrows much from a tutorial presentation by Jon Timmis, now at York.

  2. Overview • What are Artificial Immune Systems? • Background immunology • Why use the immune system as a metaphor for computation • Immune System Inspired algorithms

  3. Artificial Immune Systems • Relatively new branch of computer science • Using natural immune system as a metaphor for solving computational problems • Not modelling the immune system – too hard • What the IS does is detect invading/unusual things; • What AISs (usually) do is detect rare/suspicious events, by borrowing computational ideas from the IS • Variety of applications so far … • Fault detection (Taylor, Corne) • Computer security (Forrest, Kim) • Novelty detection (Dasgupta) • Robot behaviour (Lee) • Machine learning (Hunt, Timmis, de Castro)

  4. Basic Immunology I

  5. The Role of the Immune System • It protects our bodies from infection, operating via: - A first line nonspecific line of defence: barriers - A second nonspecific line of defence: general attack. Then comes specific (i.e. targeted) defence, comprising: • Primary immune response • Launches a response to invading pathogens • Secondary immune response • Remembers past encounters, leading to: • Faster response the second time around

  6. Basics and Terms A Pathogen is any agent (bacterium, virus, etc) that can cause us trouble THE IMMUNE SYSTEM IS OUR PRIMARY DEFENSE AGAINST PATHOGENS IT CONSISTS OF NONSPECIFIC AND SPECIFIC DEFENSES. NONSPECIFIC DEFENSES ARE THE BODY'S FIRST LINE AGAINST DISEASE. THEY ARE NOT DIRECTED AGAINST A PARTICULAR PATHOGEN. THEY GUARD AGAINST ALL INFECTIONS, REGARDLESS OF THEIR CAUSE. SPECIFIC DEFENSES ARE ATTEMPTS BY THE BODY TO DEFEND ITSELF AGAINST PARTICULAR PATHOGENS. Since Pathogens must enter the body in order to cause disease, the body's first line of defense is to keep pathogens out. So, what organ is used for this?

  7. Basics II The Body's MOST IMPORTANT Nonspecific Defense is the SKIN.  UNBROKEN Skin provides a continuous layer that protects almost the whole body.  Very Few Pathogens can penetrate the layers of dead cells at the skin's surface. Oil and sweat glands at the surface of the skin produce a salty an acidic environment that kills many bacteria and other microorganisms. The importance of the Skin as a Barrier against Infections becomes obvious when a small portion of skin is broken or scraped off: Infection almost always follows. Infections are a result of the penetration of the broken skin by microorganisms normally present on the unbroken skin. Pathogens also enter the body through the Mouth and Nose, but the body has Nonspecific Defenses that protect those openings.

  8. MUCOUS MEMBRANES are Tissues that protect the interior surfaces of the body that may be exposed to pathogens. They serve as a barrier and secret MUCUS, a sticky fluid that traps pathogens. MUCUS, CILIA, and HAIRS in the Nose and Throat trap Viruses and Bacteria. Pathogens that make it to the Stomach are destroyed by Stomach Acid and Digestive Enzymes. Many Secretions of the Body, including MUCUS, SALIVA, SWEAT, and TEARS, CONTAIN LYSOZYME, AN ENZYME THAT BREAKS DOWN THE CELL WALL OF MANY BACTERIA. But what happens if something gets past all that ?

  9. The Inflammatory Response This is the SECOND LINE OF DEFENCE When Pathogens get past skin and mucous membranes, and enter the Body, this Second Line of Defence comes into play, triggered by injury to tissues in the body. The injured cells release a protein called HISTAMINE, which starts the a series of changes called the Inflammatory Response.

  10. THE INFLAMMATORY RESPONSE IS A NONSPECIFIC DEFENSE REACTION OF THE BODY TO TISSUE DAMAGE. Histamine increases blood flow to the injured area and increases the permeability of the surrounding capillaries, as a result,  Fluid and White Blood Cells (WBC) leak from blood vessels into nearby tissue. Pathogens are attacked by PHAGOCYTES, WHICH ARE White Blood Cells THAT ENGULF AND DESTROY PATHOGENS

  11. The most common Phagocyte, 50 to 70 percent of the White Blood Cells in the body, is the NEUTROPHIL. Neutrophils circulate freely through blood vessels, and they can squeeze between cells in the walls of a capillary to reach the site of infection.  They then engulf and destroy any pathogens they encounter

  12. Another type of Phagocyte (also a White Blood Cell) is the MACROPHAGE; they consume and destroy any pathogens they encounter, they also rid the body of worn out cells and cellular debris. Some Macrophages are stationed in the tissues of the body, awaiting pathogens, while others move through the tissues and seek out pathogens.

  13. NATURAL KILLER CELLS are large white blood cells that, unlike phagocytes, attack cells that have been infected by pathogens, Not the Pathogen Themselves. They are particularly effective in killing Cancer Cells and Cells Infected with Viruses. A Natural Killer Cell punctures the cell membrane of its target cell, allowing water to rush into the cell, causing the cell to burst

  14. But if all that is not enough … IF A PATHOGEN IS ABLE TO GET PAST THE BODY'S NONSPECIFIC DEFENSES, THE IMMUNE SYSTEM REACTS WITH A SERIES OF SPECIFIC DEFENSES THAT ATTACK THE DISEASE CAUSING AGENT. This is called the IMMUNE RESPONSE A SUBSTANCE THAT TRIGGERS THE SPECIFIC DEFENSES OF THE IMMUNE SYSTEM IS KNOWN AS AN ANTIGEN. AN ANTIGEN IS A SUBSTANCE THAT A MACROPHAGE (WBC) IDENTIFIES AS NOT BELONGING TO THE BODY.

  15. The Immune Response involves several organs, as well as White Blood Cells in the Blood and Lymph. These include the BONE MARROW, THYMUS, LYMPH NODES, TONSILS, ADENOIDS, AND SPLEEN. Each organ of the immune system plays a different role in defending the body against pathogens. Bone Marrow manufactures the billions of WBC needed by the body every day.  Some newly produced WBC remain in the bone marrow to Mature and Specialize, while others travel to the Thymus to Mature. Lymph Nodes Filter Pathogens from the Lymph and expose them to White Blood Cells The Spleen, a fist-sized organ located behind the stomach, Filters Pathogens from the Blood.  It is stocked with WBC that respond to the trapped pathogens.

  16. Where is it?

  17. Self/Nonself distinction In order to Respond to Pathogens, but to avoid responding to and destroying cells from its own body, Lymphocytes MUST BE ABLE TO RECOGNIZE A PATHOGEN AS A FOREIGN INVADER AND DISTINGUISH IT FROM CELLS OF THE BODY. This is the key to it all, and where most of the inspiration comes for computational systems.

  18. The Immune Response:The last line of defence The general idea is this: Something has got through the first lines of defence, and entered the body in force. If the body has been invaded by this particular nasty thing before, then special Lymphocytes called B-Cells and T-Cells are able to recognise these specific pathogens, and overwhelm them (thanks to the `immune system memory’ If this is a new invasion, then the B-Cells will learn how to fight this invader. (and then remember for next time).

  19. Specific Antigen Recognition Nasty thing This lymphocyte recognises the red pathogen B-cell or T-cell Surface receptor molecule This one doesn’t B-cell or T-cell

  20. Generating variety The receptor molecule is a protein, encoded by a highly variable gene. There is essentially a combinatoriallibrary of parts in the genome: Each B or T cell makes up its receptor by choosing: and one of these, etc… one of these and one of these dna The result is that an enormous variety of possible surface receptors could be chosen. This is effectively a method for generating random receptors. Since recognition need not be exact, it is possible in practice for a B or T cell to generate a receptor which matches any given antigen.

  21. Generating variety II In addition, B-Cells (but not so much T cells) also undergo somatic hypermutation. Somatic just means in the body, during one’s lifetime. Hyper just means `a lot’. In a nutshell: 1. A B-cell recognises an antigen 2. A complex chain of events then leads to this B-cell dividing, creating daughters who produce the same receptor. 3. But these daughter cells may have mutations in their library. 4. Some of the daughters may recognise the antigen even better. 5. Back to 1.

  22. Clonal Selection and Negative Selection The whole process (antigen recognition, consequent production of new B-cells with similar receptors, repeated …) is called Clonal selection. In AIS paralance it is also called positive selection (you’ll soon see why). But how come the immune system doesn’t generate receptors which cause it to recognise (and hence then try to destroy) bits and pieces which are valid and necessary parts of the body? It does! But B or T cells with such self receptors get destroyed by a Process called negative selection. The standard picture (from the book by Timmis and de Castro) is on the next slide.

  23. Clonal and Negative Selection 1 2 3 4 5

  24. Clonal and Negative Selection • In the picture, we see the fate of five different B-cells, each with • A different receptor molecule. Note, these are also called antibodies. • Much simplified: • 1 & 5. These ones find themselves recognising a `self-antigen’. This leads to them • getting killed off (`clonal deletion’). This happens as part of the cell’s `schooling’. • Before release into the blood (lymph), B-cells (T-cells) are exposed to a full • range of self-antigens in the bone marrow (Thymus). They are killed if they • recognise anything. Hence, those that graduate and enter the system are only • those that will recognise foreign invaders. • 2 & 4. These find themselves going round the body a few times without recognising • anything. Thus, they are never stimulated to divide and multiply, and soon die. • Clonal expansion/positive selection: this B-Cell recognises something – the • recognition process causes it to divide, producing daughters who will have • similar, possibly higher affinity, receptors (and those with better affinity will • have more offspring, etc …). They don’t divide forever. Some become stable as • `memory cells’ (ready to fight if infected with the same pathogen again), others • become plasma cells, which secrete lots of the recognising antobody into the blood.

  25. Interim Summary A pathogen comes along: If it gets through the barriers (skin, etc), nonspecific lymphocytes kill it, as part of the `inflammation’ response in reaction to injury. If it gets past that (I.e. there’s so much of it, it gets into the bloodstream anyway), then the Immune Response comes into play, as follows: If we’ve seen this one before, there are antibodies in the blood (secreted by memory cells); these antibodies disable and/or tag the invader. The tagging attracts killer cells to make sure it is destroyed. If we haven’t seen this before, B-cells and T-cells are floating around with a great variety of surface receptors. One of these will at least recognise it a bit. Clonal expansion then happens, and with gene variability and somatic hypermutation we eventually get some B or T cells which are capable of recognising it. The associated antobodies then disable and tag the invaders.

  26. Some interesting related points Some ailments are `beyond’ the immune system, since they either directly disable it, or work faster than it, or both (or something else). Cancer: the problem here is uncontrolled growth and multiplication of normal cells. If caused by any specific pathogen (controversial) then it could be that just a tiny amount needs to go unattacked for a short time, and the problem starts. Leukaemia: a cancer of the bone marrow – it (and its treatment) throw an enormous spanner into the heart of B-cell production. Vaccination: this is where we deliberately provoke an immune response to small levels of a pathogen (or something similar to it), so that our IS is ready if there is a real infection. AIDS: some T-cells (called Helper T Cells) are the main players in most of the things we have looked at. E.g. via special messenger molecules, they activate the clonal expansion of B cells! The HIV virus directly attacks Helper T-cells, essentially disabling the immune system.

  27. AIS Algorithms The IS is the inspiration for a whole new field of computer science which is building systems, for various purposes, which borrow ideas from the workings of the IS. The basic ideas and algorithms are less easy to pin down (than with EAs, or NNs, e.g.). However, the most easily abstracted algorithms (which are also most frequently `borrowed’, are: negative selection and positive selection. The scenario is typically this: We need to detect anomalous patterns (network attacks, bank card fraud, unusual temperature/vibration/pressure patterns in machinery, etc…) The space of normal patterns is very large and variable (we can’t just say `anything which doesn’t look like X is bad’.) We have little or no idea about what anomalous patterns will look like. So, we use IS ideas: The overwhelmingly common approach is: Negative selection: generate random detectors (receptors), but filter them by testing their affinity to known self patterns. Each new pattern / window of data is then matched against these detectors.

  28. Anomaly Detection: the most common application of AIS • The normal behavior of a system is often characterized by a series of observations over time. • The problem of detecting novelties, or anomalies, can be viewed as finding deviations of a characteristic property in the system. (I.e. non-self) • For computer scientists, the identification of computer viruses and network intrusions is considered one of the most important anomaly detection tasks

  29. Negative Selection Algorithms Developing the detector set Using the detector set

  30. Basic Notes on Negative Selection Algorithms • A robust system should detect any foreign/strange activity rather than looking for specific known patterns of intrusion. • No prior knowledge of anomaly (non-self) is required • This lack of prior knowledge is useful, because we normally have very few, or no, example data sets of intruders (e.g. attacking patterns of telnet packets, fraudulent credit card use patterns), so standard classification by (e.g. ) NNs can’t be done.

  31. That’s all Use google to find out more about Artificial Immune Systems, if you wish. Generally it has not yet been a clearly successful research area, since it is not clear that successes so far could not have been achieved just as well by conventional methods. However (my personal opinion), the more complex the system that we need to protect, and the more complex and varied the potential threats, perhaps the more like natural immune systems the protection approach needs to be.

More Related