Technical Information

Schematic Overview of Anvita's architecture

Anvita is an integrated computer reference of science & medicine, including information (and tools for interpretation of data) encompassing internal medicine, molecular pathology (biochemistry, molecular biology, pathology), laboratory medicine, radiology, pharmacology, anatomy, & statistics. The Anvita strategy differs from the stategy of building separate systems for basic science & clinical medicine, then expecting them to somehow come together.

Fast, Focused, Friendly
Anvita is a local application which employs efficiently compiled code resulting in "real time" "point of contact" access to information using a combination of interactive graphics, images & text. Anvita will run without internet connection. This makes Anvita MUCH faster than an internet-based application & useable in circumstances where internet is not available (or logistically functional).

Anvita queries return specific information rather than a list of possibly relevant items. A browsing tool through selection of mouse-sensitive objects facilitates further focusing or redirecting of the initial inquiry.

Interaction with Anvita is through an intuitive "point & click" or "fill in the blank" interface. Anvita translates user input to map that to its underlying representation, initiates a search, then translates search results into a combination of interactive graphics, images & text that is presented to the user as output. Selectable items will "highlight" when the mouse is passed over it. Common usage terms are accepted as input for "fill in the blank" queries.

Knowledge Representation and Features of Intelligence
Anvita is a functional prototype addressing:
1) representation of medical & biological information,
2) presentation of data,
3) reasoning with this information.
It is written in the robust, portable object-oriented programming environment of Common Lisp & the Common Lisp Object System (CLOS) [1] & the Common Lisp Interface Manager (CLIM). This provides a single development environment, from prototyping to application delivery on multiple platforms.

CLIM links presentation of an object directly to its semantics. This facilitates the separation of the internal representation of objects from the presentation of information to users. CLIM facilitates both the translation of user input into methods manipulating Anvita objects & the presentation of the data represented in Anvita objects to the user. Thus the interface may be designed to optimize user friendliness & data presentation without compromising features of the underlying CLOS
representation.

Anvita development began as a project in knowledge representation. The initial questions addressed included:
- What are the essential features of biomolecules that give rise to their function within cellular compartments & individual cells?
- How can these features be captured in a computer representation?
- What are the features of cells that give rise to their interactions with other cells, within tissues, organs & organ systems?
- What are the features of diseases that are essential to understand for the practice of medicine & how are those features related to changes in organ system, organ, tissue, cellular, compartmental & molecular function?

The questions addressed in knowledge representation must also accept that uncertainty will always be associated with scientific data. The uncertainty may be quantitative or qualitative in nature. Qualitative uncertainty shares representational issues with generalizations [2], the distinction being especially important when reasoning with information.

The object-oriented paradigm of the Common Lisp Object System (CLOS) [6] provides a means of generating structures of arbitrary complexity & also a means of organizing information within a classification structure. The CLOS classification within Anvita
serves several functions, including:
- structure for representation of biological information
- inheritance of properties common to similarly classified objects
- customized operations through methods defined on specific classes
- representation of qualitative uncertainty & generalization
- entry points for specification of queries

Extensive use of object-oriented techniques has been employed in the development of Anvita [1,2,3,5]. These include object-oriented user interface techniques & the metaobject protocol [4]. The metaobject protocol provides a framework for the system to examine its own behavior (introspection) & modify it to achieve a desired effect. This facilitates development of an application tailored for knowledge representation in multiple domains.

Objects have slots which provide a means of describing an object in detail. Slots are specialized descriptions of objects defined with the most generalized class to have that attribute or property. For example, molecule has slots "compartment" & "size"; molecular complex has the additional slot "subunits". Slots may assume default values specified with the class definitions & most slot values are themselves objects or lists of objects. Slots & their default slot values are inherited through the classification structure [4]. The inherited slot default values may be accepted as is or may be further specified. Classes of objects may have multiple supertypes so their properties may be determined by inheritance of multiple slot default values.

Slot default values provide a means of programming biological knowledge into Anvita. Slots are defined in Anvita as logical "AND" or logical "OR" slots. For example, the slot "compartment" is a logical "OR" slot, in that a single molecule may be located in the cytoplasm or nucleus, but not both at the same time. In contrast, the slot "motif" is a logical "AND" slot. Proteins contain structural elements that give rise to function(s) of the molecule. Most proteins consisting of a single polypeptide can be represented as an ordered set of motifs connected by peptide regions. These peptide regions may be further organized into domains. This representation of proteins facilitates the representation of molecular interactions. Genes are represented through similar considerations.

One of the useful features of Anvita is the display of molecular structure. Small molecules are simply drawn from methods defined on specific classes of molecules. Proteins are drawn similarly using methods defined on specific classes of proteins. These methods use symbolic representations & generate line drawings that illustrate protein motifs & domains containing the motifs. The drawings themselves contain selectable regions to show further detail. Proteins containing multiple transmembrane domains are also drawn using similar but more complex methods, taking into account loop size, cytoplasmic & exoplasmic domain size & intramolecular bonds across loops. Molecular complexes are drawn using methods that align subunits & draw intra- & intermolecular bonds. Genes are drawn using methods that show structural features of the 5'enhancer region, 5' promoter region, introns & exons & 3' region. Methods for drawing gene clusters have also been implemented.

Small molecules, proteins, molecular complexes & macromolecular complexes participate in molecular events, non-covalent bindings & chemical reactions (enzymatic & template-directed). These events may be linked in series to form molecular pathways or cascades. Molecular pathways cross-talk with other pathways in signaling networks. Anvita contains structures to hold detailed information about molecular events & molecular pathways & to display this information in a graphical format. Display of molecular events is facilitated by methods that draw the events from symbolic representations of the substrate(s), product(s) & (if indicated) enzyme & template. Molecular pathways are drawn on an individual basis.

Sometimes it is known that a molecular event interacts with another, but the details of that interaction are not known. For example, phosphorylation of one protein may result in translocation of another to the nucleus (mediation) or inhibition of enzymatic activity of another protein phosphorylation (modulation). This type of information is often useful & Anvita provides CLOS structures for capturing this information & utilizing it in analyses of signaling pathways.

Much of the structural framework for simulating signal transduction networks [6] is in place, representation of molecules, events (including kinetic parameters) & molecular interactions. Issues of generalization & uncertainty need special attention in the analysis of signaling pathways. Qualitative uncertainty is represented by using a sufficiently general class to describe a molecule. Generalization is represented by using the most general class of molecule to participate in a particular event (exceptions may be specified). A general class is interpreted by Anvita to represent uncertainty, unless otherwise specified. Generalization is distinguished from uncertainty during the comparison of objects.

Anvita Use Case Example
An 80 year old veteran presents to the VAMC accompanied with spouse with chief complaint of memory impairment. This is a teaching clinic, a medical student and a resident in internal medicine and a fellow are in clinic, three individuals with significantly different knowledge bases. The medical student wants to know how formulate an assessment and plan. The resident wants to know about how to distinguish among the different neurodegenerative disorders (tauopathies, synucleinopathies, polyglutamine diseases, and other neurodegenerative diseases, including vascular dementia, prion diseases and normal pressure hydrocephalus). The fellow wants to know about investigational therapies for Alzheimer’s disease.

Anvita contains guidelines for evaluation and management of thousands of disorders linked to and integrated with their molecular bases when known. For the medical student, the Anvita dementia entry is organized similar to other entries according to etiology, epidemiology, pathology, genetics, clinical manifestations (including elements of the history and physical), laboratory, special laboratory (includes procedures), radiology, differential diagnosis, complications, management and references. Pharmaceutical agents within the management section are organized according to indications, contraindications, dosage (including dosage adjustment in renal failure, if indicated), pharmacokinetics, adverse effects, drug interactions, mechanism of action and references. Virtually any item of medical significance in an Anvita window is selectable, so that the student may explore for example relevant laboratory tests or elements of the differential diagnosis. Anvita also has and entry within the table of contents under geriatrics for office-based geriatric assessment, a good starting point for medical students.

Anvita is organized within a classification structure. Thus, the resident may compare neurogenerative disorders using Anvita’s browsing mode. The epidemiolology, pathology and clinical manifestation, laboratory, special laboratory and radiology sections of the various neurodegenerative disorders serve to distinguish the different disorders.

The management section of the Anvita entry for Alzheimer’s disease contains an entry for investigational therapies for Alzheimer’s disease. Investigational therapies are included within Anvita for many diseases. Each is associated with the molecular bases of the therapy if known. In the case of Alzheimer’s disease, the fellow may be interested in non-steroidal anti-inflammatory drugs (NSAIDs) that also inhibit APP-gamma secretase that cleaves the amyloid-precursor protein within its membrane region to generate A-beta.

REFERENCES:
1) Steele GL. Common LISP the Language. 2nd ed. Digital Press 1990
2) Ball SS, Mah VH. Senex: CLOS in molecular pathology. Uncertainty, generalization, and the comparison of objects. In: Proceedings of the 4th Annual Lisp Users and Vendors Conference; 1994 August 15-19; Berkeley, CA.
3) Keene SE. Object Oriented Programming in Common LISP, Addison-Wesley, Reading MA, 1989.
4) Kiczales G. The Art of the Metaobject Protocol, MIT Press, Cambridge MA,1991.
5) Gu H, et al. Representing the UMLS as an object-oriented database: modeling issues and advantages. Journal of the American Medical Informatics Association 2000;(7):66-80.
6) Lindvall JM, Blomberg KE, Smith CI, In silico tools for signal transduction research, Brief Bioinform 4(4):315-24, 2003