|Home | About | Journals | Submit | Contact Us | Français|
Knowledge of multiscale mechanisms in pathophysiology is the bedrock of clinical practice. If quantitative methods, predicting patient-specific behaviour of these pathophysiology mechanisms, are to be brought to bear on clinical decision-making, the Human Physiome community and Clinical community must share a common computational blueprint for pathophysiology mechanisms. A number of obstacles stand in the way of this sharing—not least the technical and operational challenges that must be overcome to ensure that (i) the explicit biological meanings of the Physiome's quantitative methods to represent mechanisms are open to articulation, verification and study by clinicians, and that (ii) clinicians are given the tools and training to explicitly express disease manifestations in direct contribution to modelling. To this end, the Physiome and Clinical communities must co-develop a common computational toolkit, based on this blueprint, to bridge the representation of knowledge of pathophysiology mechanisms (a) that is implicitly depicted in electronic health records and the literature, with (b) that found in mathematical models explicitly describing mechanisms. In particular, this paper makes use of a step-wise description of a specific disease mechanism as a means to elicit the requirements of representing pathophysiological meaning explicitly. The computational blueprint developed from these requirements addresses the Clinical community goals to (i) organize and manage healthcare resources in terms of relevant disease-related knowledge of mechanisms and (ii) train the next generation of physicians in the application of quantitative methods relevant to their research and practice.
Practicing physicians and surgeons leverage knowledge of pathophysiology mechanisms to take clinical decisions. While the physiological community1 generated this knowledge as a result of precise quantitative analysis, the pathophysiology-based methods employed by the clinical community for the training and practice of medicine primarily rely on qualitative approaches (given, for instance, the heavy reliance on free-text descriptions and static diagrams in pathophysiology textbooks, as well as in clinical record-taking). This divergence in approaches between the Physiome and Clinical communities is at the root of considerable loss of opportunity to share knowledge and collaborate on the study of pathophysiology mechanisms relevant to:
To allay the collaborative handicap caused by the above discrepancy, it is crucial:
In addressing points A–C above, this paper identifies and collates the representational requirements for the exchange of computer-readable knowledge about pathophysiology mechanisms between the Physiome and Clinical communities. This blueprint of requirements builds a shared view of pathophysiology mechanisms—a view that is consistent and compatible with the operational goals of both communities. In the conclusion section, we make recommendations for this blueprint as the basis of an open computational toolkit for the management of pathophysiology resources in tangible support of cross-disciplinary collaboration.
Knowledge of multiscale anatomy is leveraged by both Physiome and Clinical communities to organize knowledge about:
Both communities develop biophysical models, to a varying degree of formality, that take into account correlations between biomedical measurements. In some cases, it is important to understand how the route linking the locations of measurement contributes to the correlation. In addition, mathematical depictions of biophysical mechanisms, in both normal physiology and pathophysiology, may be represented in terms of rates of transfer over these routes.
Furthermore, both communities take into account the creation, destruction and alteration of routes as leveraged by the body to regulate rates of transfer over a wide range of physical dynamic systems, for instance in:
The implications of these obstacles are discussed in the next section.
Considerable effort has been invested by
However, the above investment has a somewhat limited effect on the consistent bridging of (i) the physiological meaning of mathematical models of processes to (ii) the physiological meaning of histological, radiological and healthcare models.
Multiscale anatomical knowledge of measurement location, and routes linking such locations, is not formally described and shareable between the two communities. Developing a formal topological representation of biological structure across multiple scales to address the above shortcoming is a considerable challenge. While, for instance, substantial investment by the systems biology community has gone into the clear cataloguing of well-characterized compendia of relevant biological structures (i.e. proteins , small molecules  and subcellular locations ) that participate in molecular mechanisms, the unmanageable size of the combinatorial space bearing every possible anatomical compartment has severely curbed the ability of any single structural model of multiscale anatomy to enumerate them all. The challenge for the Physiome and Clinical communities remains, therefore, to provide the means to systematically map the anatomical location of a wide range of measurements. The following scenarios indicate some of the shortcomings to be surmounted.
To allay the above shortcomings, the next section focuses on the step-wise depiction of a specific pathophysiology mechanism to elicit the representational requirements for the recording of (i) the location of measurements and (ii) the routes over which processes responsible for the correlation of these measurements unfold.
The first part of this section describes the biological steps in the pathophysiology mechanism of hydronephrosis caused by calculus obstruction of the ureter. In the second part, this biological description is converted into a graph linking the locations of measurements, relevant to the evolution of this pathology, over an anatomical route that accounts for the correlation between measurements. Generalized representational requirements for a database of pathophysiology pathways are discussed in the third part.
There are a number of conditions  that lead to urinary solutes precipitating out of solution to create calculi, for instance:
The calculus formation, therefore, increases the likelihood of further stone accretion by stimulating conditions (d) and (e) above. The growth of a stone in the pelvicalyceal region of the urinary tract provides the right conditions for a calculus to reach a size that cannot be subsequently conveyed down the ureter . Such an accretion becomes lodged at the pelviureteric junction, reducing urinary outflow. This flow reduction is followed by a build-up of upstream pressure as the nephrons in that kidney continue to produce urine. This urinary pressure build-up goes on to compress the blood supply of the kidney in the hilum, leading to vascular strangulation and subsequent atrophy of that organ.
The above informal account of the pathophysiology mechanism of hydronephrosis implicitly describes a number of correlations of rate and state measurements drawn from a range of locations along and across the renal epithelial (i.e. urinary tract) and endothelial (i.e. blood vessel) conduit systems. Therefore, we look to the basic geometric configuration of these conduit systems to motivate the organization of structural knowledge about the location of these measurements. The basic organizational features that need to be taken into account to represent routes linking measurement locations are that:
The type of biomedically relevant measurement in the above scenario is of two kinds: the state property of some Material (M) or the rate property of some Process (P). At the very least, therefore, there are four distinct types of located measurement entities, symbolized as follows: PW, PC, MW, MC.
In this pathophysiology scenario, correlations of located measurements can be biophysically modelled over the six equations below (numbered I–VI). These functional relationships define transfers taking place within C, within W or across CW, as follows:
(I) The changing composition of urine; precipitation of salts
The concentration of various biochemical substances in urine, such as calcium, urate, phosphate, etc. (Ci, i = 1,2…), changes with time at rates
dependent on various biochemical processes. At some point, an ion such as calcium (concentration C1) supersaturates and precipitates out of solution, at which point a kidney stone begins to form.
(II) Formation of calcified material
The mass of calcified material, mCa, accumulates at a rate
dependent on the rate of decrease of calcium in solution (−dC1/dt) and other factors such as inflammation (I, e.g. due to infection or irritation), which causes mucus proteoglycans to be excessively secreted into urine.
(III) Changing flow in the ureter
Flow in the ureter, qureter, depends on the diameter of the ureter at the point of stone formation (dureter), the mass of the kidney stone (mCa), the filtrate flow rate (qfiltrate) and the viscosity of urine (v):
If the stone grows large enough to block the ureter, the flow will stop, but the consequences of stone formation are felt upstream well before this point is reached.
(IV) Distension of the renal calyces and the renal pelvis
The difference between filtrate flow (qfiltrate) and flow in the ureter downstream of the stone (qureter) is absorbed by distension of the renal calyces and the renal pelvis (which have a total volume Vr):
The fluid pressure in this space (pr) depends on both the volume Vr and the elasticity of the surrounding tissue:
(V) Pressure in the surrounding tissue
The pressure pt in the surrounding tissue depends on the pressure pr in the renal calyces and the renal pelvis and on the elasticity Et of the tissue:
(VI) Compression of the renal vasculature
The increasing pressure in the surrounding tissue pt compresses the renal vasculature and increases the resistance to arterial blood flow qa, which also depends upon geometric and material properties Ea of the vascular wall and the pressure pa in the blood:
The above relations provide the topology of a graph of C- and W-located M and P measurements depicting the key types of transfer relevant to the pathology of hydronephrosis. A schematic illustration of this graph is shown in figure 3. The direction of the arrows in this figure indicates the implied causal influence from r.h.s. to l.h.s. of equations I–VI. This graph also provides two examples of long-range transfers, namely (figure 4):
The edges of the above graph represent pairwise transfers from one location to another. To determine the route between the two locations, however, an independent computer-readable topological model of kidney structure is required to determine if, at the level of granularity of this model, the two locations are contiguous. If the two locations share a border according to the structural model, then the route for transfer consists of the union of the two contiguous locations. If there is no contiguity between the two sites, then path-finding calculations over the structure of kidney conduits is necessary to determine path through other contiguous locations that constrain the transfer.
So far, we have described a graph of transfers relevant to the mechanism of a single pathology. On a more general level, point A above outlined a spectrum of transfer scenarios drawn from normal physiology as well as disease. As this spectrum of physical dynamic systems are also amenable to interpretation and representation in terms of energetic operations, the provision of a coherent computational blueprint to record these graphs is a key first step to developing a resource of physiology and pathology mechanisms shared by the Physiome and Clinical communities. The development of a Resource for Pathophysiology Mechanism (RPM) needs to address the following high-level requirements in its blueprint:
In practice, the technical solution may take the form of an application program interface (API) for core methods that articulate and maintain the topological model, such that calls to this API can be embedded within specialist tools serving either community.
The biomedical meaning of electronic health record (EHR) content is organized using disease terminologies such as SNOMED-CT [35,36]. One avenue of semantically bridging the biomedical meaning of Physiome models with that of EHRs is to develop a map between disease terms and variables in process models. The hydronephrosis (SNOMED-CT ID D7-14106) scenario is just one example of the definition of a standard disease entity in terms of a transfer graph of located measurements. While the manual curation of a single disease mechanism is achievable, the coherent coverage of pathology mechanisms over the entirety of the SNOMED-CT disease terminology requires a community collaborative effort, together with the appropriate tools to effect such collaboration. The requirements collected in this work set out the blueprint of such a toolkit. The implementation of such a blueprint will provide a key means for the Clinical community to explicitly contribute and collect its knowledge about pathophysiology mechanisms to improve the integration of Physiome models for training and healthcare decision support.
The implementation of the blueprint will be discussed in our future work.
1The Physiome community is distinct from the physiological community. The Physiome community has developed modelling and data standards, and associated computational tools and repositories, to create a reproducible multiscale physiological modelling framework. We will refer to both the physiological community and the Physiome community from here on as the ‘Physiome’ community.
We declare we have no competing interests.
The authors gratefully acknowledge grants from the Innovative Medicines Initiative (IMI) grant nos. 115156-2 (DDMoRe) and 115568 (AETIONOMY), as well as European Union FP7 grant agreement 600841 (CHIC).