Novel approach to clinical data analysis: application to kidney transplantation

Lead Research Organisation: University of Warwick
Department Name: Sch of Engineering

Abstract

The research aims to develop a novel mathematical approach to build a model of antibody pathogenicity in antibody incompatible kidney transplantation (AIT). Currently, renal replacement is expensive and cost-effective provision of kidney transplantation constitutes a major global health priority. The number of patients receiving renal replacement therapy exceeds 1.4 million worldwide and is growing by 8 percent annually, in excess of the growth rate of the general population. Solutions for the prevention or reversal of renal disease have so far failed to significantly change the development of global patient numbers. An economically viable alternative to allograft transplantation is not foreseen in the mid-term future; hence renal transplant is the only solution.

The successful outcome of a transplantation depends on how well the donor and recipient are matched for tissue proteins called HLA. Since only about 25 percent of transplants can be fully matched many needing a kidney have developed antibodies against HLA and these can cause transplant rejection. Patients with preformed HLA antibodies wait longer or cannot receive a kidney.

AIT has been pioneered making it possible to reduce levels of antibody before surgery and transplant 'mismatched' patients. More than 40 percent of kidneys are however still rejected. This is because complete elimination of antibodies is not possible. Types of harmful antibodies and levels to which they must be reduced are also not known. Traditional clinical studies utilise standard statistical analysis that requires very large number of participants and have until now failed to predict kidney rejection. The project will therefore employ an alternative approach that combines statistical analysis with the development of novel methods of dynamic patterns analysis.

The human immunological reaction to kidney transplantation will be modelled in the framework of non-linear stochastic systems approaches followed by their translation into clinical context via following objectives: (1) to develop an appropriate methodology and subsequently analyse dynamic and static properties of antibody evolution in AIT (2) to provide the most informative data processing tool for antibody risk assessment (3) to create a rigorous foundation for a comprehensive data source based on the available research data. The research aims to address the following clinical questions: (a) what types of preformed HLA antibodies are most harmful and associated with significant risk of kidney rejection; (b) what are critical levels of preformed antibodies at the time of surgery i.e. how much of the antibodies can be tolerated to ensure safe acceptance of the donor kidney?

This engineering and primarily non-medical strategy has not been attempted before. The strength of this project lies in its translational aspect from areas of maths/engineering to tackling medical challenges thus strengthening the cross-disciplinary Biomedical Engineering interface. This is concomitant with the unique clinical data set available for the project. The key aspect of the data concerns the patients being sampled daily in the critical first 3-4 weeks following transplantation when antibody levels change rapidly thus capturing key events in the behaviour of antibodies. This is the only programme anywhere in the world to have carried out with such level of antibody monitoring.

For the long term perspective, the project outputs will bring significant clinical benefits through an in-depth understanding of the humoral immune response in AIT. Also, this will improve risk management of transplants and the provision of access to transplantation for many untransplantable patients. The database created will consolidate research activities in this emerging field and advance outputs of the research of national and international significance.

Planned Impact

The research will bring benefits to the following areas and individuals:

Society: Clinical outcome of incompatible kidney transplantation will be improved by providing access to transplantation for many patients who are currently considered to be untransplantatble for end stage renal failure. Increased number of safely transplanted patients will consequently reduce waiting times for operation thus enhancing quality of life of people with failed kidneys. In the long term, identification of the rejection risk in individual patients based on the analysis will have a major clinical impact offering tailored immunosuppression and improved risk management.

Economy: Currently, renal replacement is expensive and cost-effective provision of kidney transplantation is a major global health priority. Worldwide, the number receiving renal replacement therapy is estimated to exceed 1.4 million and grows at a rate far more than the growth rate of the general population. An economically viable alternative to allograft transplantation is not foreseen for the near to mid-term future as solutions targeting the prevention of the disease have thus far failed to significantly change the development of global patient numbers. Therefore, renal transplant is the only solution. The new knowledge generated from this research will identify factors associated with significant risk of kidney rejection and make it possible to safely transplant patients with reduced rejection risk. This will avoid unnecessary treatments or further costly operations incurring lower treatment costs and will increase bank of kidney donors.

Knowledge: This collaborative project will produce cutting-edge research maintaining unique world leading position of the UK in two specific areas: nonlinear systems and antibody incompatible renal transplantation (AIT). Three fundamentally AIT related novel aspects of this research are highlighted: (1) Dr D Zehnder and Dr R Higgins (UHCW) have pioneered AIT in the UK and UHCW is the largest centre for AIT in Europe; (2) the antibody level monitoring in Prof D Brigg's laboratory (NHSBT) is unique and comprises the only programme anywhere in the world thus providing necessary 'dynamic' data set to develop and validate new modelling methodologies; (3) nonlinear stochastic systems approach will be developed by the PI for the first time to model pathogenicity of antibody.

People: The project will provide professional development for a PhD student with transferrable comprehensive skills in statistical and dynamic analysis of complex systems, and their application to a clinical problem. The student will gain diverse skills such as technical writing, presentation, management and research dissemination. The student will benefit from training in a clinical laboratory and networking within a multidisciplinary environment including Warwick Engineering in Biomedicine (WEB) and Science City Renal Research Group. PI will also train, supervise and mentor undergraduate and postgraduate students at MSc level.
The project will present the PI with a timely opportunity to reinvigorate and further her academic career after a career break. She will be able to participate, as an equal and independent researcher, in highly focussed and mutually beneficial collaborative activities and consequently establish herself as an authority in this evolving field of Biomedical Engineering research of both academic and clinical significance.

Publications

10 25 50

publication icon
Khovanova N (2015) Neural networks for analysis of trabecular bone in osteoarthritis in Bioinspired, Biomimetic and Nanobiomaterials

publication icon
Khovanova N (2015) Subclass analysis of donor HLA-specific IgG in antibody-incompatible renal transplantation reveals a significant association of IgG4 with rejection and graft failure. in Transplant international : official journal of the European Society for Organ Transplantation

publication icon
Lowe D (2015) Meeting report: 3rd international transplant conference: how much risk can you take? in International journal of immunogenetics

publication icon
Zhang Y (2016) A data driven nonlinear stochastic model for blood glucose dynamics. in Computer methods and programs in biomedicine

 
Description It was known that donor specific antibodies (DSA) represent a risk factor for early transplant rejection and influence allograft survival times. The aim of this research was to investigate the role of all DSA IgG subclasses (1-4) in the immune response in order to identify any potentially damaging antibodies and their influence on short and long postoperative outcome. 120 transplanted samples were available for analysis.
Firstly, we have demonstrated that one specific class of Immunoglobulin G (IgG4) was predictive of acute kidney rejection and it associated with poor graft survival. Thus, pre-operation IgG4 antibody level assessment may be used as a biomarker for risk stratification in kidney transplantation
Second major result is that we calculated the harmful levels of these IgG4 antibodies. For this we utilised Machine Learning, in particular Decision Trees and Random Forest methods, which has not been done before.
Third major outcome is the development of a dynamic model describing the behaviour of Immunoglobulin G antibodies after kidney transplantation. The model is proved useful in classification between two clinically different groups of people - who experienced episodes of acute antibody mediated rejection (AMR) of kidney in the first 30 days after operation, and who did not. This approach is found useful in capturing properties of antibody evolution from their peak concentration to final settling level. A higher frequency of oscillations and a faster antibody dissipation rate for the AMR group had been observed. The findings have important implications for the development of laboratory assays that might define the nature of the mechanisms responsible for the falls in DSA levels post-transplant, since a fuller understanding of these mechanisms might allow for pre-transplant manipulation of DSA levels and improved clinical outcomes. This is particularly important with respect to the oscillating nature of DSA levels, which may reflect a system slowly reaching homeostasis, and may be reflected in laboratory measurements.
Furthermore, in 2017 and 2018, we expanded upon this research by applying a clustering technique to identify the smaller groups of DSAs and their respective association with acute antibody mediated rejection, thus heading towards personalised approach to assess and identify the set of DSAs and their dynamics that might cause the rejection with subsequent graft loss. Our results, considered 133 DSAs from a single medical centre, have indicated the presence of two major groups in which DSA fall in to: fast rise and subsequent steep fall, and slower rise and slow fall. Collectively these groups account for approximately 60% of the DSAs, with other smaller groups comprising 40%. Our results indicate that a faster rise and subsequently steeper fall correlate stronger with the likelihood of AMR, which complements previous research in this area.
One other major recent finding is on the times of DSA monitoring in the first 50-60 most critical days after transplantation. We recommend that kidney transplant patients with both class I and class II DSA who experience an acute AMR episode should be monitored between 12-33 days after their rejection episode for best indication of a potential graft failure. A threshold of 10,000 MFI could be applied around day 20 post AMR to predict chronic graft failure in those with DSA levels greater than 10,000. Assessing individual DSA in a patient can also be indicative of graft failure but the differences are less prevalent and demand more investigations.
Exploitation Route In engineering: novel methodology has been developed to assess risk factors for medical applications with small and wide data sets.

In clinical practice for pre-transplant antibody risk assessment: IgG4 can be additional biomarker for risk stratify kidney transplant recipients.

Furthermore, in clinical practice for early post-transplant antibody dynamics assessment: IgG donor specific antibody levels could be monitored at specific time points in order to predict the risk of kidney failure.
Sectors Healthcare

URL http://www2.warwick.ac.uk/fac/sci/eng/staff/nk/research/biomed_research/
 
Description The grant finished in September 2015; however, it generated a lot of collaboration in the sector of kidney transplantation between the School of Engineering and Clinical research centres. Collaboration with Leeds Teaching Hospital and enhanced collaborative activities with University Hospitals Coventry and Warwickshire (UHCW) resulted in invited talks for clinical professionals, for example, on various mathematical approaches relevant to the area of kidney transplantation. Jointly we have produced new knowledge on harmful types of antibodies, which goes beyond the particular research task covered by the EPSRC project. We continue applying the methodology developed during the grant execution period to discover new antibody properties and their relation with negative (and positive) transplantation outcomes. The engineering team is now comfortably hosting a session on Mathematical Modelling at the Kidney Transplant International Workshop organised every two years at Warwick University in collaboration with UHCW. Moreover, the methodologies developed during EPSRC-funded project has been successfully applied by our team outside kidney transplantation area - scientific community working in the area of control of diabetes has benefited from predictive models developed recently by our group.
First Year Of Impact 2016
Sector Healthcare
Impact Types Economic

 
Description PhD scholaship
Amount £30,000 (GBP)
Organisation University Hospitals Coventry and Warwickshire NHS Trust 
Sector Public
Country United Kingdom
Start 03/2016 
End 03/2019
 
Description PhD scholaship
Amount £28,000 (GBP)
Organisation University of Warwick 
Sector Academic/University
Country United Kingdom
Start 03/2016 
End 03/2019
 
Description Vice Chancellor fully funded International PhD scholaship
Amount £80,000 (GBP)
Organisation University of Warwick 
Sector Academic/University
Country United Kingdom
Start 09/2013 
End 03/2017
 
Title A new data-driven model for post-transplant antibody dynamics in 
Description A data-driven model in the form of a third order differential equation has been developed to describe the antibody dynamics after kidney transplantation for the first time. The model is based on a unique dataset of thirty nine antibody time series from patients with and without acute antibody mediated rejection episodes. The model successfully captured the common features of the measurement time series across the cohort, and is proved useful in classification between two clinically different gro 
Type Of Material Computer model/algorithm 
Provided To Others? No  
Impact 1. Development of laboratory assays that might define the nature of the mechanisms responsible for the falls in donor specific antibody (DSA) levels after kidney transplantation. 2. A fuller understanding of these mechanisms might allow for pre-transplant manipulation of DSA levels and improved clinical outcomes. 3. We have already shown that the subclasses of IgG are associated with clinical outcomes, so that measuring the levels of these subclasses at more time points might be valuable. 
 
Title Artificial neural networks for small data sets in biomedical domain 
Description Biomedical systems are often characterised by small datasets due to the complexity and high costs of experiments on living tissues. In the biomedical domain, Machine Learning methods for regression modelling of small datasets (less than 10 observations per predictor variable) remain scarce. We addressed the problem of sporadic fluctuations and validation issues that appear in regression Neural Networks (NNs) trained on small datasets. This has been done by developing a novel framework comprising of a method of multiple runs for model development and surrogate data analysis for model validation. The significance of this work: the novel methodology is not constrained to a particular medical application and provides a general framework for application of regression neural networks to small datasets. 
Type Of Material Computer model/algorithm 
Provided To Others? No  
Impact Despite their superior performance, accuracy and versatility, neural networks are generally viewed in the context of the necessity for abundant training data. This, however, is rarely feasible in the biomedical domain, where the size of datasets is constrained by the complexity and high cost of large-scale experiments. To the best of our knowledge, effective strategies for regression tasks on small datasets have not been considered; this necessitated the establishment of framework for application of neural networks to small datasets in biomedical engineering and clinical data analysis. We expect that our model will be useful to many scientists who apply machine learning methods to medical data set for risk stratification or clinical outcome prediction. 
 
Title Significant association of IgG4 with rejection and graft failure in kidney transplantation 
Description Machine learning application for small data sets. Developed Decision Tree model which allowed to identify harmful types and levels of antibodies in high risk incompatible renal transplantation 
Type Of Material Data analysis technique 
Provided To Others? No  
Impact Pretreatment Immunoglobulin 4 (IgG4) donor specific antibody (DSA) levels correlated independently with higher risk of early rejection episodes and medium-term death-censored graft survival. Thus, we have proposed to use pre-treatment IgG4 DSA as a biomarker to predict and risk stratify cases with graft rejection and failure. Further investigations are needed to confirm our results. 
 
Description Blood and Transplant Birmingham 
Organisation NHS Blood and Transplant (NHSBT)
Country United Kingdom 
Sector Public 
PI Contribution Same contributions as for Kidney Transplant Group (UHCW)
Collaborator Contribution Supply data for analysis and help to interpret results of modeling
Impact Same outputs as with Kidney Transplant Group (UHCW). Clarification: three of our groups (my, UHCW and HNS Blood and Transplant) form one research team and should be considered together.
Start Year 2013
 
Description Kidney transplantation group 
Organisation University Hospitals Coventry and Warwickshire NHS Trust
Country United Kingdom 
Sector Public 
PI Contribution Mathematical modeling for antibody risk prediction in incompatible kidney transplantation. By analysing data and performing mathematical modeling, we discovered a particular type of antibody (IgG4), which is harmful for transplanted kidneys and causes acute kidney rejection as well as chronic kidney failures. We also estimated dangerous concentrations of these antibodies. This has not been known before.
Collaborator Contribution Our collaborators care for the patients, transplant kidneys, collect data following surgical operations and help us to interpret results of modeling and data analysis.
Impact 1. Harmful types of antibodies in incompatible kidney transplantation for both acute graft rejection and chronic graft failure 2. Levels to which the antibodies should be reduces for safe kidney transplantation. 3. Novel dynamic model for immunoglobulin G antibodies demonstrating the differences in the evolution of the antibodies after transplantation in those people who accepted and those who rejected the organ. Collaboration is multidisciplinary: nonlinear physics/dynamic systems and immunology.
Start Year 2013
 
Description St. James's University Hospital, Leeds Teaching Hospitals NHS Trust 
Organisation Leeds Teaching Hospitals NHS Trust
Country United Kingdom 
Sector Public 
PI Contribution Hospital supplied data from patients. Our group, in turn, performs predictive modelling and analysis for risk stratification, and develops novel methodologies to cope with data of limited sizes
Collaborator Contribution Contribution with data and analysis.
Impact This collaboration is multi-disciplinary and includes areas of immunology and predictive modelling
Start Year 2016
 
Description 26th Annual BSHI Conference, Cambridge 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Poster presentation, networking with medical doctors. Received best poster presentation award.
Year(s) Of Engagement Activity 2015
URL http://www.bshi.org.uk/BSHI_Conference_Diary_Dates.pdf
 
Description 4th International conference on transplantation at University of Warwick: organising committee member and session chair 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact We attracted transplant doctors and researchers to this conference, confirmed our leadership in the area of research, increased collaborative opportunities, developed plans for multi institutional studies within the UK and beyond.
Year(s) Of Engagement Activity 2016
URL https://www.mededcoventry.com/Courses/Renal/conference.aspx
 
Description American Transplant Congress 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Two oral presentations at American Transplant Congress. This congress proved huge success as there were just a few academics, like us, with non-medical background to present our mathematical models and results to the clinical doctors. Useful discussions with opportunities to collaborate across institutions and perform multi-institutional studies.
Year(s) Of Engagement Activity 2015
URL http://2015.atcmeeting.org/
 
Description Annual British Transplantation Society Conference, Glasgow 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact 3 presentations delivered. Networking with clinical doctors.
Year(s) Of Engagement Activity 2016
URL https://www.bts.org.uk/BTS/Events_Meetings/Annual_Congress/BTS/Events_Meetings/Congress_2016/Annual_...
 
Description Conference presentation 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Clinical Relevance of Complement Activating Pre-formed Donor Specific Antibodies in Crossmatch Positive HLA-incompatible Renal Transplantation: a Multicentre Study. Luncheon Session - Immucor at 32nd European Immunogenetics and Histocompatibility Conference. Venice, Italy, 9-12 May 2018.
Year(s) Of Engagement Activity 2018
 
Description Conference presentation 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact S Daga, R Higgins, N Khovanova. Immunological risk stratitification pre-transplantation and risk of early antibody mediated rejection - analysis of UK antibody incompatible transplant registry data. Annual British Transplantation Society Conference. Brighton, UK. 14 - 16 Mar 2018.
Year(s) Of Engagement Activity 2018
 
Description Conference presentation 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Complement Activating anti-HLA Class 2 antibodies are associated with poor renal allograft survival: multicentre study. American Transplant Congress. Seattle, Washington. 2-6 Jun 2018.
Year(s) Of Engagement Activity 2018
 
Description ENOC 2014 - 8th European Nonlinear Dynamics Conference, Vienna 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? Yes
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Conference publication (short paper) by the PhD student.

Training for the PhD student in the area of nonlinear stochastic systems modelling.
Year(s) Of Engagement Activity 2014
URL http://enoc2014.conf.tuwien.ac.at/index.php/welcome
 
Description Grand round: invited talk 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Professional Practitioners
Results and Impact Educational talk for clinical professionals on mathematical modelling for clinical applications. The aim was to demonstrate the opportunities which modern machine learning and advanced engineering can open up, compare with the traditional methods of medical statistics, in (a) predicting outcomes of clinical interventions and (b) improving clinical risk stratification.
Year(s) Of Engagement Activity 2016
 
Description International Conference on Biomedical and Health Informatics, Valencia 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? Yes
Geographic Reach International
Primary Audience Policymakers/politicians
Results and Impact The talk sparked discussion and I was congratulated on an impressive presentation at the BHI 2014 conference.

Based on the reviews of the conference paper and presentation, the work was invited to submit an expanded version of the conference paper for publication in the IEEE Journal of Biomedical Health Informatics (IEEE - J-BHI). J-BHI is one of the leading journals in computer science and information systems with a strong interdisciplinary focus on biomedical and health applications.
Year(s) Of Engagement Activity 2014
URL http://bhi.embs.org/2014/
 
Description Invited Talk 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Professional Practitioners
Results and Impact Machine learning for data analysis in the area of organ transplantation. Invited talk. Leeds Teaching Hospitals NHS Trust. Leeds, 11 October 2017.
Year(s) Of Engagement Activity 2017
 
Description Invited presentation: discussion of strategies for collaboration with Oxford University 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Professional Practitioners
Results and Impact Sharing experiences between two working groups. The aim was to demonstrate the opportunities which modern machine learning and advanced engineering can open up, compare with the traditional methods of medical statistics, in (a) predicting outcomes of clinical interventions and (b) improving clinical risk stratification.
Year(s) Of Engagement Activity 2016
 
Description Invited talk 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Invited talk. Role of C3d assay in risk stratification. Immucor Europe Meeting, Stockholm, 21 Mar 2018.
Year(s) Of Engagement Activity 2018
 
Description Joint British Transplantation Society and Nederlandse Transplantatie Vereniging Congress, Bournemouth 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? Yes
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact 2 poster presentations delivered. Discussions with potential collaborators and contributors to the database, which is being developed as part of the project

Potential collaboration with Johns Hopkins (USA)
Year(s) Of Engagement Activity 2015
URL http://www.bts.org.uk/MBR/Educational/Archives/Congress_Abstracts/iSamples/Member/Educational/Congre...
 
Description Member of organizing committee and conference presenter 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? Yes
Geographic Reach International
Primary Audience Other academic audiences (collaborators, peers etc.)
Results and Impact Extremely useful conference which provided training in the area of transplantation and gave an opportunity to share with clinicians the novel bioengineering approaches of data analysis and modelling.

Positive feedback, numerous useful discussions afterwords and suggestion to collaborate. We hope to unite our efforts to make antibody incompatible transplantation safe by creating a joint database with larger sample sizes in order to be able to develop mathematical models suitable for clinical implementation and use. Further collaborative activities are expected with University Hospitals Coventry and Warwickshire, and with Guy's and St Thomas Foundation Trust - London.
Year(s) Of Engagement Activity 2014
URL https://www.mededcoventry.com/Courses/Renal/conference.aspx
 
Description Scientific Congress 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Poster presentation, networking activities
Year(s) Of Engagement Activity 2017
URL https://www.bts.org.uk/BTS/Events_Meetings/Annual_Congress/BTS/Events_Meetings/Congress_2016/Annual_...
 
Description Symposiaum on Biological and Medical Systems (Berlin) 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact 2 oral presentations were delivered by my 2 PhD students. Both talks obtained top marks and were invited for journal publications. Collaborative activities discussed with the group from Uppsala University.
Year(s) Of Engagement Activity 2015
URL http://www.bms2015.org/
 
Description Third International Transplantation conference, Warwick University 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? Yes
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Talk sparked a lot of questions and discussions and positive feedback. It has lead to further collaborative activities (with University Hospitals Coventry and Warwickshire, and with Guy's and St Thomas Foundation Trust - London). New ideas have been discussed on the usefulness of decision trees and artificial neural networks approaches for small data sets in clinical practice.

Extremely useful conference which provided training for the PhD student in the area of transplantation and gave an opportunity to share with clinicians the novel bio-engineering approaches of data analysis and modelling. The PhD student was involved in the technical committee of the conference.
Year(s) Of Engagement Activity 2014
URL https://www.mededcoventry.com/Courses/Renal/conference.aspx