Can a novel mathematical system for animal models help solve this dilemma?

Big data has become a commonly used term in many walks of life. In the context of biomedical research, big data refers to the extremely large experimental or clinical data sets that are computationally analysed to reveal patterns, trends and associations.

Radiotelemetry is a commonly used technique to monitor continuous physiological data in conscious animals. Small devices are surgically implanted into lab species to record how their bodies respond to drugs or disease.

These devices capture waveform data such as blood pressure, electrocardiograms (ECGs),or respiratory flow. The data is typically sampled at 1000 Hz – that’s 1000 numerical data points every second. This equates to 86,400,000 data points in one day – which is, quite frankly, a few too many numbers for the average scientist to handle. Instead, we tend to focus on specific parts of the waveform such as the maximum, minimum, rate, amplitude and – in the case of ECGs – intervals. These simplified averages are displayed to the end user for analysis.

By simplifying the waveform data in this way, most of the available numerical data is being discarded – data which pertains to the shape and variability of the waveform. It has been known for some time that the shape of an arterial pulse wave (or ECG) contains important information, and qualitative examination of these waveform shapes can give us more clues about the performance of a system. However, qualitative analysis relies on a trained individual continually viewing the waveforms, and this is simply not practical.

A research meeting organised by NC3Rs in 2013 brought together biomedical scientists, who pitched a data challenge to mathematicians. During this meeting, mathematician Philip Aston applied different mathematical models to radiotelemetry data from pharmacologist Manasi Nandi’s laboratories at King’s College London. This collaboration led to a brand new way of quantifying the shape and variability of repeating waveforms (Aston et al., 2018; Nandi et al, 2018).

The figure below demonstrates how a periodic wave can be mathematically transformed into a three dimensional shape to allow quantitative assessment of all the collected data.

  • Periodic data with three points identified on the wave from a fixed point (Z, Y, and X)
  • X, Y, and Z values are plotted in 3D.
  • Three points on a single wave translate to a triangle shape in the Attractor analysis
  • Data from many hundreds/thousands of waveforms are overlaid to form the Attractor. Data from the orientation, angles, and lengths of the sides can all be used to inform structural changes to the system being measured.

Figure: Attractor reconstruction. Click here for a video of this method.

Using the method to answer important questions

One of the projects that recently arose was a chance collaboration between Dr. Mary McElroy and Dr. Aileen Milne based at Charles River Labs in Edinburgh. Mary and Aileen hosted Carolyn Lam, an industry MSc student from King’s College, who had previously worked with Dr. Nandi as an undergraduate. Carolyn used the attractor to analyse respiratory wave forms from a rat model of fibrosis. Preliminary data demonstrates that fibrosis is associated with a dramatic rounding of the attractor shape, while the normal triangle shape is partially restored in a treatment group (Abstract accepted for the ATS 2019).

Big data analysis using the Attractor has enormous potential to sensitively detect changes in many body systems, including the respiratory systems, to better model and predict the efficacy and safety of new drugs and to bridge the gap between preclinical and clinical studies.

References:

Philip J Aston, Mark I Christie, Ying H Huang, Manasi Nandi Beyond HRV: attractor reconstruction using the entire cardiovascular waveform data for novel feature extraction Physiol Meas. 2018 Feb; 39(2).

Manasi Nandi, Jenny Venton, Philip J Aston (2018) A novel method to quantify arterial pulse waveform morphology: attractor reconstruction for physiologists and clinicians Physiol Meas. 2018 Oct; 39(10).

McElroy MC, Lam C, Cranston I, Baily J, Aston P, Nandi M and Milne A. Respiratory Assessment Using Intrapleural Pressure.Head-out Plethysmography in a Repetitive Bleomycin Challenge Model of Pulmonary Fibrosis in Conscious rats.  ATS abstract, 2019

AL solutions for earlier disease detection organised by the BIA https://www.youtube.com/watch?v=c4VJGlFDN-8

(Charles River scientists Mary McElroy and Aileen Milne, and Carolyn Lam of Kings College London also contributed to this blog post)