List of questions
How can we reduce bias when using data to train a machine learning system?
We have built an app, called CogniCare, which supports carers of people with dementia. Carers can register to use the app and provide information about themselves, and the person they care for. They can also track symptoms over time. On their home feed, carers receive suggestions personalised based on this information. We are now building an intelligent engine using machine learning to build a model for personalising information based on the data we receive from the carer, findings from research, and existing datasets. We want to reduce bias as much as possible, but require expert input on what the best approach would be.
What are the best models for personalisation and predictive algorithms?
We have built an app, called CogniCare, which supports carers of people with dementia. Carers can register to use the app and provide information about themselves, and the person they care for. They can also track symptoms over time. On their home feed, carers receive suggestions personalised based on this information. We want to use self-generated data and data from research to build models using machine learning that use this data to personalise suggestions and provide predictive care.
What analytical platforms are available that will enable direct detection of proteins as well as nucleic acid biomarkers at the same time from the same biological sample in the same machine, especially if taking a single ‘liquid biopsy’ sample for clinical and research purposes?
At the moment there has been a significant development of platforms for ultrasensitive detection, such as the Simoa platforms of Quanterix Corporation, the Mecwins Scala Bio, The Ziplex of ANGLE Biosciences, the SMCx-Pro platform of Merck Millipore. There is also the well-established bead based Luminex Magpix and Flexmap platforms in the marketplace. However, the ability to perform same time, same sample, same machine ‘combo-assays’ to detect both proteins and nucleic acid biomarkers has been limited by insufficient platform sensitivity to perform direct detection of nucleic acid biomarkers from biological samples. Up to now, multiple extraction/isolation/PCR amplification steps have been required, making combo-assays not a practical opportunity. The ability to simultaneously monitor levels of both protein and nucleic acid biomarkers would be highly valuable, particularly if this could be performed on a single machine and in the same time frame. It would increase the quality of date, as well as the sample processing performance and efficiency in-lab. This has benefits for academic research, pre-clinical and clinical trials in drug development, and could deliver a real breakthrough for patients post clinical procedure interventions, monitoring patient health/relapse as well as for drug resistance build-ups. Undertaking additional research and development with emphasis on integrating the nucleic acid detection side more closely with the classic methods and machines used for protein detection would be valuable. Direct detection chemistries for nucleic acids capable of delivering results free of ‘false positives’ through error free reads to single nucleobase resolution would be valuable for integration here.
What analytical methods can be developed to enable better analysis of circulating tumour DNA (ctDNA)?
Improved methods to isolate and quantify circulating tumour DNA remains challenging, but is an important element for successful ‘personalised medicine’ to emerge. Blood samples are easy to collect, minimally invasive compared to classical tissue biopsies, so there is a real appeal for patients and clinician alike, including post-operative monitoring where multiple tissue biopsies may not be a realistic option. Various challenges remain, including the low concentration of ctDNA found in blood, their relatively short lives once in the bloodstream, individual cancer patient levels, and relationship between cancer stage and ctDNA quantified, leading to contradictory/error prone results. There is a need to develop more robust protocols and understand the similarities and differences inherent in the different analytical platforms now in use, such as Illumina, Roche Cobas and reagents used,in order to arrive at valuable ‘point-of-care’ assays for cancers. An industry wide approach linked to independent academic research would be valuable.
How can we nudge adaptive health behaviours such as sleep, activity and better nutrition using ubiquitous technology (e.g. phones) and artificial intelligence?
Detalytics is a human data analytics platform that helps people understand the influence of daily lifestyle choices on their overall condition. Knowing is not enough though, people also need to change their habits to genuinely feel better and increase their life quality
What approach would you recommend to build capability in accessing machine level data and then to turn this information into real time actionable insights?
We have a large manufacturing footprint with a complex mix of equipment of varying ages. Currently we have varying degrees of access and connectivity to the PLC data from this equipment and therefore limited insights are generated from it. We would like to build our strategy in this space to improve our machine level connectivity, generate more data, understand the best way to store this data and then subsequently visualise and analyse it to help us generate insights which we can action in real time. We’d like to move beyond descriptive analytics and into the predictive and prescriptive space. This could be to support performance improvements on the lines or areas like predictive maintenance for example.
How do you build digital capability and culture across a large, diverse, manufacturing and supply chain business?
We have a large number of employees working in a number of functions in our business, covering areas such as innovation, procurement, production, planning and logistics. The workforce is multi-generational with varying capability and needs for digital skills. As we look forward, we see that new technology and digital will be a key enabler in driving our business forward. We would like to understand how best to approach building the capability of our workforce in digital to ready them for the introduction of new technology in their roles. This could be from building basic IT skills, through to more advance analytics capability and on to data science and programming skills. We also feel that it’s key we build the right culture in the organisation where employees embrace digital and the change this may bring. This could mean creating a data driven culture for example, where everyone understands the importance and power of data and the insights we could generate from it and places this at the heart of our operation.
How do you consolidate data provided in different formats to make it analytics ready?
Instrumentation from differing vendors provides data in varying formats which has to be assembled and transformed into a common format that can be interrogated by model machine learning techniques. This is often a slow and arduous process and results in an Excel spreadsheet. How can this process be streamlined to removed the manual intervention and possibility for error.
How can we replace two on-line Gas Chromatograph analysers with a system that can infer the results that the analysers provide by using continuous process data?
Methyl methacrylate (MMA) is a monomer used in the manufacture of poly methyl methacrylate (PMMA), or acrylic polymer products as they are more commonly known. Acrylic products are common and used for a very diverse range of applications including paints and coatings, optical systems, signage, kitchen worktops, prosthetics, etc. Lucite International, wholly owned by the Mitsubishi Chemicals Corporation, is a global leader in the design, development and manufacture of acrylic based products. Lucite is always striving to improve its process technology. One improvement area relates to the control of a reaction step, where currently the chemical components are measured in the feed and exit to the reactor using an on-line Gas Chromatograph. The GC application is difficult because of the composition of the process streams and the need for fairly complex sample conditioning systems. Lucite would like to explore the possibility of replacing on-line analysers in this duty with an alternative means of providing the same results. The project is at an early stage, and as yet we have not come up with many solutions. We would like to investigate the potential to correlate multiple process data inputs with the composition of the output streams and would like to collaborate with others who have up-to-date knowledge and experience in this area.
What would be the best way to create a development dashboard to automate parts of the analysis of spatial datasets?
Marine Scotland Science has access to a variety of spatial datasets. We often receive request from the public to access this information. Due to sensitivities, we are required to undertake repetitive analysis in order to anonymise and aggregate data. A dashboard allowing us to automate (parts) of this process will be great.
How can we use data to increase adherence to prescribed treatments amongst young people with asthma and to provide insights into how behavioural change might affect this?
There are on average 3 people dying each day in the UK as the result of an asthma attack, with two of these deaths preventable if the individual had been adhering to their asthma management plan. Targeting young people with asthma as they transition from children’s services to adult services as they deal also with the other major changes that are happening in their lives at this time, we are working towards creating an industry led “triple helix” open innovation collaboration that will seek to develop innovative solutions that will improve adherence amongst this patient cohort, the learning from which we will then use to spread to other age groups and other long term health conditions. The collaboration will be formed with partners from health and social care, a range of academics, the third sector and SME’s working alongside at least one large lead multi-national company, based here in Edinburgh. The aim being to bid for Innovate UK monies and other research funding sources to deliver life changing innovative solutions, that will have spread globally improving the lives of the 339 million people world-wide who have asthma, the prevalence of which is increasing.
How might we use Artificial Intelligence to increase the speed of analysis so that our clients may quickly make decisions to build and/or digitize marine infrastructure assets?