genetic pathways的计算机模拟

[编者的话]

Kastner发表了名为“Modeling a Hox Gene Network in Silico Using a Stochastic Simulation Algorithm”的文章，是一个非常有启发意义的工作，下文是对该工作的简要评论。

The early development of the vertebrate embryo requires a greater level of transcriptional activity than at any other stage in the life cycle. The specification of numerous cell types in the correct spatial and temporal pattern demands a very fine control over this transcription. As a result, the overall complexity of genetic interactions in development are vast, and might always defy a complete algorithmic description. However, a number of developmental processes occur relatively independently of others, and this has allowed them to be modelled by computers, a process referred to as in silico modelling.

Jason Kastner and his colleagues at the California Institute of Technology have used this approach to study the genetic interactions that help pattern the early hindbrain. Development of the vertebrate hindbrain is relatively well understood, and involves a physical subdivision into a number of segmental compartments, or rhombomeres. Each rhombomere expresses a distinctive set of transcription factors, membrane receptors and ligands that define its identity and determine its function in the adult brain. Kastner and his colleagues have concentrated on two rhombomeres in particular, r4 and r5. Embryological and genetic experiments in mice and chicks have revealed that these rhombomeres express the transcription factors Hoxb1 and Krox20, respectively. Hoxb1 expression is maintained in r4 via a cofactor dependant, autoregulatory loop, and Hoxb1 is thought to repress Krox20 expression. A further transcription factor, Hoxb2, is required in both r4 and r5. Its expression is maintained by different mechanisms in each rhombomere, requiring Hoxb1 in r4 and Krox20 in r5.

The authors have used a stochastic algorithm to mimic these genetic interactions in silico. This algorithm incorporates the known promoter and enhancer elements in Hoxb1, Hoxb2 and Krox20, and uses starting values that represent the conditions at a very early stage in hindbrain development. It also takes into account cell division and random changes in gene expression known to occur in a few cells at early developmental stages.

It is very encouraging that the results of this analysis match closely those obtained in live embryos. This also holds true with experimental manipulations of gene expression. For example, a Hoxb1-/- background can be simulated by simply removing Hoxb1 from the algorithm. In this case Hoxb2 expression is lost from r4, something which is known to occur in real Hoxb1-/- embryos. Intriguingly the in silico simulation also suggests that the Krox20 gene is strongly expressed in r4 of Hoxb1-/- embryos, a novel finding that still awaits experimental confirmation in vivo.

The in silico model also predicts the 'misfiring' of a small number of cells, whereby they stop expressing the developmental genes of their neighbours and deviate from their normal fate. The model shows that if this occurs at an early stage in hindbrain development some cells can recover and express the right genes again. This recovery is not possible if it occurs at a later stage though, and the cells fail to adopt any specific developmental fate. The success of this simple in silico model is highly encouraging, and suggests that this approach could yield new biological insights.

by Richard Morgan

相关文章请见：

Kastner J et al. (2002). Modeling a Hox Gene Network in Silico Using a Stochastic Simulation Algorithm. Dev Biol, 246:122-131