<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-0907">
  <Title>Marking Time in Developmental Biology: Annotating Developmental Events and their Links with Molecular Events</Title>
  <Section position="3" start_page="0" end_page="46" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Most relation extraction work to date on biomedical articles has focused on genetic and protein interactions, e.g. the extraction of the fact that expression of Gene A has an effect on the expression of Gene B. However where genetic interactions are tissue- or stage-speci c, the conditions that govern the types of interactions often depend on where in the body the interaction is happening (space) and at what stage of life/development (time).</Paragraph>
    <Paragraph position="1"> For genetic pathways involved in development, it is critical to link what is happening at the molecular level to changes in the developing tissues, usually described in terms of processes such as tubulogenesis and epithelialization (both involved in the development of the kidney) and where they are happening.</Paragraph>
    <Paragraph position="2"> The processes themselves are usually linked to stages rather than precise time points and spans like 6.15pm EST , March 3 , last year .</Paragraph>
    <Paragraph position="3"> Within the developmental mouse community, there are at least two different ways of specifying the developmental stage of an embryo - Theiler stages (TS), and days post coitum/embryonic day (d.p.c./E). However, these cannot be simply mapped to one another as can days, weeks and years. Embryonic days are real time stages independent of the state of the embryo and dated from an assumption about when (approximately) the relevant coitus must have taken place, while Theiler stages are relative stages dependent on the processes an embryo is undergoing.</Paragraph>
    <Paragraph position="4"> Developmental stages can be also be referred to implicitly, by the state of the embryo or the processes currently taking place within it. This is because during development, tissues form, change, merge or even disappear. So if the embryo is undergoing tubulogenesis, one can assume that its developmental stage is (loosely) somewhere between TS20 and birth. If the text refers to induced mesenchyme during a description of tubulogenesis, one can assume that this change in the mesenchyme is the (normal) consequence of the Wolffian duct invading the metanephric mesenchyme.</Paragraph>
    <Paragraph position="5"> The invasion is known to occur around 10.5 d.p.c so the induced mesenchyme must come into existence soon after this time.</Paragraph>
    <Paragraph position="6"> Temporal links between developmental events may be indicated explicitly (e.g. first, a tubule develops into a comma-shaped body, which then develops into an S-shaped body), but they are more likely to be indicated implicitly by their ordering in the text and by associative (or bridging ) anaphora where the anaphor refers to the result of a previously mentioned process, e.g. the induction of metanephric mesenchyme as one event, and a subsequent mention of induced mesenchyme (an associative or bridging reference) within another event, suggesting the former event occurred before the latter.</Paragraph>
    <Paragraph position="7">  ney morphogenesis. The arrows show directed interactions between genes that are required for the speci ed processes. E.g Pax2 interacts with (activates) Six2 which, together with Sall1 and Wt1, is required for differentiation of the mesenchymal cells in the metanephric mesenchyme. Image taken from (Ribes et al., 2003).</Paragraph>
    <Paragraph position="8"> This work on linking molecular and developmental events mentioned in text on development is also meant to deal with the problem that no one article ever fully describes a topic. The partial genetic interaction network in Figure 1 has been built from several different studies and not determined from just a single experiment. So not only does the information within one article need to be mined for useful information - the information across articles needs to be associated with each other with respect to temporal, spatial and experimental grounding. Eccles et al. (2002) states that Pax2 is required for differentiation of mesenchymal cells during kidney morphogenesis, while Sajithlal et al. (2005) states that Eya1 is required. However these two results by themselves do not help us determine whether these requirements are independent of one another or whether they are required at different stages or in different parts of metanephric mesenchyme or whether the two genes interact. The conditions involved in the experiments, most importantly the temporal conditions, can help to link the two events.</Paragraph>
    <Paragraph position="9"> This work aims to develop methods for extracting information from text that will ground genetic pathways (molecular events) with regard to tissue location, developmental process and stage of embryonic development - that is, their spatio-temporal context. The task at hand is to recognise how biologists write about developmental events and then adapt existing or formulate new natural language processing techniques to extract these events and temporally relate them to each other.</Paragraph>
    <Paragraph position="10"> The resultant information can then be used both for database curation purposes and for visualisation, i.e. to enrich pathway diagrams such as Figure 1, with information such as when and where the interactions take place, what type of interactions are involved (physical, activation, inhibition), the origin of this information and other associated information.</Paragraph>
  </Section>
class="xml-element"></Paper>