2nd Interdisciplinary Workshop On Artificial Immune Systems

26th to 27th April 2005

Following on from the success of last years workshop, ARTIST are pleased to announce a second workshop on interdisciplinary research in Artificial Immune Systems.

The workshop starts at 11am on Tuesday 26th April and finishes at 4pm on Wednesday 27th April

The first day will be consist of presentations and discussion forums in the areas of immunology, mathematical modeling of the immune system and computation modeling of the immune system.

The second day will be focused around the use of these techniques in AIS, with a view on applications of AIS in the wider context. The speakers, and timetable, will be announced shortly.

Places are limited, so please register early. ARTIST will cover the cost of the workshop, one nights accommodation and all meals etc from the Tuesday lunchtime until the Wednesday lunchtime.

Travel costs will NOT be met by ARTIST.

Invited Talks:

Innate and adaptive immunity

R. A. Robins

Abstract: Innate immune receptors are germline encoded, subject to evolutionary pressure to recognise previously encountered specificities. Adaptive immune recognition uses randomly generated receptors, allowing the recognition of completely novel specificities (antigens). The adaptive system also has memory, allowing rapid recognition on second encounter with antigen. The advantages of this flexible but randomly encoded adaptive response also brings the problem of how to distinguish self from non-self, and make an appropriate response in each case. Interactions between the innate and adaptive immune systems are increasingly being recognised as critical for this essential control of the immune response. Examples of these innate/adaptive interactions will be discussed, to indicate how the quality and quantity of the adaptive response may be largely governed by the innate immune response. A deeper understanding of the interactions between receptor systems with randomly generated receptors and those with preprogrammed specificity will be necessary to create effective artificial immune systems.

Safety in randomness

H. A. van-den-Berg

Abstract: The immune response protects our bodies against pathological influences. Detecting those influences is a primary task. A human data engineer readily conceives of this task as being delineated in a fairly straightforward manner. It would, for instance, be natural to regard it as some sort of data classification problem. However, the immune system achieves detection (and subsequent control of morbidity) by a confluence of the activities of various, quite distinct components,some of which serve to classify the nature of an influence while others determine its identity. The latter task falls to a repertoire of detectors whose activity is controlled by signals emanating from components that determine the nature of any given influence. The detectors are generated at random, without an built-in prejudice as regards the structure of pathological influences. Moreover, the data actually recognized by the detector is not in itself directly related to the morbidity of the pathological influence; in fact, the detectors are capable, in principle, of detecting data associated with non-pathological influences as well. Only one thing is required of data recognized by detectors: that they enable the immune response to target the pathological influence with sufficient specificity. This latter property I call `saliency'. I will argue that the need for an efficient use of resources sets up a pressure to operate with minimally salient data, and that efficacy of the immune response requires `filtering': a reduction step between the totality of data in the system and the data visible to the detector repertoire. I will discuss how minimal saliency might be determined, and how minimally salient `chunks' are to be selected from larger data structures for the purpose of presentation to the detectors. Again, I will argue that this filtering has to be essentially random.