Applicability domain tool

A way to gain information on the reliability of a prediction generated by the DSS is by means of an applicability domain tool capable of estimating the applicability domain from the training set using a suitable statistical approach. Once the applicability domain has been established, it is then possible to assess whether the compound of interest falls within dehned boundaries. 

The objective of IMPROVE to develop an integrated process and product model, spanning the application domain, tool functionalities and software platforms with both a product data as well as an engineering workflow perspective, has been by far too visionary and forward-looking to get the chance of being picked up in the short-term by the software industries. 

The tools for in silico toxicology are broadly applied in the drug development process. The particular use of the tools is clearly context-dependent, which includes the quality of the prediction and the applicability domain of the model. 

There are a growing number of tools to assess applicability domain, and a number of expert systems, for example, TOPKAT and MultiCASE, have their own measures of fit. These need to be developed and their application to larger drug libraries demonstrated. 

Several attempts were performed to determine the accuracy of in silica prediction tools developed for lipophilicity (for a recent review, see [34]). The main factor limiting the accuracy of all predictive methods is the training sets used to generate the models, in terms of population and quality of the experimental data they contain. Since most of the methods proposed in commercial software were built with data available in the public domain, their accuracy can be expected to be comparable. Thus, in order to select the most suitable prediction tool, other criteria than accuracy have to be used such as the speed of the calculation for large databases, the price of commercial software or the application domain of the model. 

On the whole this approach has not (yet) delivered widely used predictive tools, certainly not those that can be used in isolation and/or by nonexperts, but some efforts in this respect have evolved into a number of commercially available systems whose predictive abilities were reviewed a few years ago [39], Probably the biggest issue with any of these methods is that their applicability domain tends to be restricted to the chemistry basis on which they were originally developed and, of course, that they are not well suited to the assessment of complex mixtures, such as botanical extracts and other materials whose chemistry is not precisely defined. 

The support of the design process by tools should be as close as possible to these application models. This facilitates that the process is carried out within the design environment and not outside of it. On layer 2, we, therefore, find external models of tools for different kinds of users, where the presentation of these tools corresponds to the languages and methods used within the design process. Analogously, the user interfaces fit the application domain, the organization of a process, its context, and the like. 

The current state of design processes can essentially not be improved by making only small steps. Instead, a new approach is necessary. Thereby, we face principal questions and nontrivial problems. We find new questions and corresponding problems by coherently and uniformly modeling the application domain and by defining new and substantial tool functionality. The layered process/product model is a scientific question which - even in a long-term project like IMPROVE - can only be answered partially. 

In this paper, we present incremental integrator tools (integrators) which are designed to support concurrent/simultaneous engineering and can be tailored to integrate any specific interdependent documents from any application domain. The only restriction is that documents have to be structured such that a graph view on their contents can be provided. 

Adaptability An integrator tool must be adaptable to a specific application domain. Adaptability is achieved by defining suitable integration rules and controlling their application (e.g., through priorities). In some cases, it must be possible to modify the rule base on the fly. 

In each application domain, e.g. in chemical engineering, there are a lot of applications for integrator tools. As a consequence, the realization of a specific integrator tool has to require as little effort as possible. 

The user interface of KomPaKt only gives a unified way of accessing synchronous and asynchronous communications. This interface integrates some useful existing communication tools. For the considered application domain, also the development of some new tools was necessary. These tools are described in this subsection. 

Adaptability. Management tools have to be adapted to a specific application domain and they must provide domain-specific operations to their users. 

Trying to optimize the runtime performance of the tools in the development environment, requires an adequate Quality of Service (QoS). QoS in the network can be implemented with IntServ, DiffServ or MPLS [1068]. However, in the scenarios considered here, a host s performance usually plays a greater role for the overall performance. Moreover, no QoS management is available in the application domain considered. There are, however, other, more flexible approaches, e.g. the application of mobile agent technology to network and system management, evaluated for example in [557] and [1001]. 

Based on the aforementioned basics of polymer compounding and process analysis the focus will now be on the formal integration issues within the CRC and its scenario. All the presented use cases were improved by the requirements from the application domains and vice versa. For example the discussion and formalization of the interfaces between plastics and chemical engineering lead to simulation tools enhancements and new features which themselves showed effects on the work processes within the domain and to external partners. 

Moreover, as tools are provided by different vendors, they are based on different system platforms, proprietary document formats, and conceptual models of the application domain. So, there is a heterogeneous landscape of existing tools. These tools constitute a proven solution for carrying out a certain activity. Engineers are familiar with these tools their use can be best practice in the application domain. Maintenance as well as further development is guaranteed due to established vendors, wide deployment, and an actively pursued dialogue between vendors and their clients. For there reasons, it is economically not feasible [973] to replace these tools by newly built ones. 

We already discussed that the PPM should be structured into different layers, from application domain models to platform models (cf. Fig. 1.6). On every layer, there is a complete description of the product, i.e. the result of the design process overall configuration). Trivially, we find hierarchies on every layer. In top-down direction, from layer to layer, the number of details is growing. The model is transformed from layer to layer, from an explicit model to more implicit descriptions (code). The product is complemented by the associated design proce.s.s on every layer. Like the product, the process is hierarchically structured. An abstract process model is transformed into tool commands and corresponding code. 

In Sect. 1.1, it was also discussed that one characteristic of IMPROVE is that we started, with application problems and models. In particular, we did not build tools which are evaluated afterwards and are either appreciated or rejected. Instead, we studied industrial work processes first to see what kind of support is necessary. Thus, we first fixed application domain models and the required functionality of tools, before building these tools. 

Application Domain Models Products, Processes, Tools... 

With respect to notations, methods, and tools we also have some partial results in this book. For example, we find the C3 notation for work processes, notations for representing application domain models, or PROGRES as type-and instance-based notation for conceptual modeling of tools. Furthermore, we find methodologies for using these languages. We also find tools and tool adaptations to support product and process modeling. 

Abstract. The first vertical column of the layered process/product model (PPM) addresses the direct, experience-based support at the technical workplaces of designers. More specifically, we demonstrate the transition from application domain models to executable tool models, focussing on the process perspective of the PPM. This vertical column is jointly realized by the Al subproject, providing the fine-grained application domain models, and the BI subproject, dealing with their conversion to executable tool models to be used by process-integrated tools. In this contribution, we provide an outline of the cooperation results. 

In this section, we describe the transition from application domain models to executable tools. This transition forms the first vertical column (b) of the layered process/product model (PPM) shown in Fig. 6.1. Two subprojects of IMPROVE are involved The Al subproject addresses the analysis of the design processes in the IMPROVE reference scenario (cf. Sect. 1.2.2) and the provision of fine-grained work process and decision models (Sects. 2.4 and 2.5). These models serve as a starting point for the development of tool support employing the process integration mechanism developed by the BI subproject (Sect. 3.1). 

The section is structured as follows First, we give an overview of the application domain models of Al and the executable tool models of BI, respectively (Subsect. 6.2.2). We successively describe the transition from application domain to tool models in Subsect. 6.2.3. Then, in Subsect. 6.2.4, we relate application and tool models to the overall layered PPM. Finally, we conclude with a summary and a discussion of still open issues (Subsect. 6.2.5). 

Prom Application Domain Models to Tool Models... 

The first vertical column of the layered PPM addresses the refinement of application domain models to tool models, the latter serving as a specification for a process-integrated environment based on process-integrated software tools. In the sequel, we describe the basic steps of this transition. 

Abstract. The models developed within subprojects A2 and B2 together form one of the vertical columns of the process/product model. The application domain models of A2 are rehned to tool models of B2 such that integrator tools can be realized. The process of building integrators is rather well understood in general, as is the process of refining the application domain models of A2 to tool models of B2. Nevertheless, important parts are missing for a concise and layered process/product model. 

In this section, we present the transition from application domain models to executable integrator tools (cf. Sect. 3.2). The transition represents a part of the overall process/product model (PPM), namely the vertical column from subproject A2, supplying the application domain models, to subproject B2, importing these models and refining them to executable specifications for integrator tools. The main focus is on the product perspective of the process/product model, as integrators deal with products of the development process. 

---