Crick Idea to Innovation (i2i)

Lead Research Organisation: The Francis Crick Institute

Abstract

Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.

Technical Summary

The Confidence in Concept scheme is a key part of MRC’s translational research strategy and provides annual awards to institutions, to be used flexibly to support the earliest stages of multiple translational research projects. The award can be used by the institution to support a number of preliminary-stage translational projects. The projects supported should aim to provide sufficient preliminary data to establish the viability of an approach –– before seeking more substantive funding.  It is intended to accelerate the transition from discovery research to translational development projects by supporting preliminary work or feasibility studies to establish the viability of an approach.
 
Title Next generation automated ultramicrotomy for high throughput array tomography 
Description The positioning of a sample in an ultramicrotome is a 4-axis alignment problem: a 3-axis rotation (pitch, yaw, roll) to align the plane of the block face to be parallel with the 2D plane defined by combination of the downward cutting sweep and the orientation of the knife, followed by a single-axis translation to overlap both planes. In existing systems, the rotational degrees of freedom are all manually controlled via thumb screws. During alignment, the translational degree of freedom is controlled manually via an electronic interface. The alignment is performed iteratively by monitoring a range of visual cues, for example the angle of a reflection of the diamond knife on the sample surface. During this process, the user often adjusts magnification and focus of the binocular viewing optics several times. Once aligned, the sample is automatically cut into a series of ultrathin sections by repeatedly sweeping the sample over the stationary edge of a diamond knife and advancing by a fixed amount. The sections float onto a boat of water behind the knife where they are mechanically pushed across the water surface by subsequent sections emerging from the knife. From here, in standard operation, they are collected manually by physically manipulating them onto a grid for mounting in the microscope. Alternatively, for higher throughput, they are collected on an automatic tape-collecting ultramicrotome (ATUMtome) system, where the sections are drawn up onto a continuously moving flexible tape on a reel-to-reel mechanism. The new Delmic imaging system will require sections to be collected onto a special substrate that is rigid and therefore not compatible with the ATUMtome. Since the ultramicrotome should be able to automatically collect many thousands of sections, it is important that any automatic handling systems are able to scale to this requirement. Alignment approach 1: The first approach will be to mimic and attempt to replicate the procedure of a skilled user by integrating computer vision and motorisation on an existing commercial ultramicrotome. This will involve building the electronics and software, along with custom adapters to mechanically couple the system to the ultramicrotome. The computer vision system will view the process through the same imaging system as the user, feeding the digital images into custom image processing software which can be used to drive the optimal adjustments via the actuators. This approach will utilise the already existing degrees of freedom of the system (2-axis adjustment via the sample mount, 1-axis adjustment via the knife mount and translation of the sample along the axis of the mounting arm). Alignment approach 2: Instead of mimicking the heuristic procedure used by an expert, alternative sensors and actuators may be incorporated, for example a multi-axis piezo stage and a laser based alignment system. This type of approach would bypass the need for human or computer vision by using a control systems approach to optimise the alignment via direct measurement and feedback. Furthermore, if time permits, with the addition of a further translation axis, this type of system could operate as a standalone ultramicrotome, perhaps for integration into a combined sectioning and imaging system. Section collection: Automated section collection onto a series of rigid substrates will involve several complex operations, including robotic manipulation of many separate delicate objects and interaction with samples on the surface of the water, with the associated effects of surface tension and fluid dynamics. A mechanism to monitor the progress will also be important in order to prevent loss of samples in the event of an error. 
Type Of Material Improvements to research infrastructure 
Year Produced 2017 
Provided To Others? Yes  
Impact Technological developments in electron microscopy (EM) have resulted in systems and workflows that can produce datasets well into the terabyte regime for a single sample. This allows us to interrogate samples from a range of health and disease studies, for example cancer, malaria, tuberculosis and neurodegenerative diseases, at high resolution in order to answer vital ultrastructural questions. By automatically acquiring data on the teravoxel scale we are able to image unprecedented volumes at the few nanometre scale. One of the most significant bottlenecks remaining is the sample sectioning process, where ultrathin slices (50-200 nm) are cut with a diamond knife on an ultramicrotome system. There is currently no fully automated system to align, cut the sample and collect the resulting sections, instead requiring painstaking manual operation by a user with many months or preferably years of prior training and experience. Many of the downstream imaging and analysis problems are caused by effects relating to this manual interaction, for example lost areas of sample due to initial misalignment and physical distortions of the sections due to variations in the manual collection process. Our proposed solution is to work with an SME (Delmic) to develop a new fully automated method of aligning the sample and collecting the sections for subsequent imaging in a high-throughput volume EM workflow, array tomography. 
 
Description Caroline Hill - AZ collaboration - Validating potential therapeutic targets in cancer 
Organisation AstraZeneca
Country United Kingdom 
Sector Private 
PI Contribution Joint research on validating potential therapeutic targets in cancer
Collaborator Contribution Joint research on validating potential therapeutic targets in cancer - AZ provision of drug expertise, technologies and libraries. Specifically AstraZeneca's phage display platform. Kinetic binding assays using the Octet platform from Hyvönen lab Cambridge.
Impact Follow on funding from i2i to continue the project.
Start Year 2017
 
Description Caroline Hill - AZ collaboration - Validating potential therapeutic targets in cancer 
Organisation University of Cambridge
Country United Kingdom 
Sector Academic/University 
PI Contribution Joint research on validating potential therapeutic targets in cancer
Collaborator Contribution Joint research on validating potential therapeutic targets in cancer - AZ provision of drug expertise, technologies and libraries. Specifically AstraZeneca's phage display platform. Kinetic binding assays using the Octet platform from Hyvönen lab Cambridge.
Impact Follow on funding from i2i to continue the project.
Start Year 2017
 
Description Early clinical study with Barts Hospital looking at the use of MRI as a prognostic tool for leukaemia 
Organisation Barts Health NHS Trust
Country United Kingdom 
Sector Public 
PI Contribution Establishment of a MRI-based approach to evaluate the bone marrow vascular functionality as prognostic biomarker in acute myeloid leukemia
Collaborator Contribution Expertise of two consultants and MRI time.
Impact Early clinical study
Start Year 2018
 
Description Edgar Deu collaboration with GSK - Pharmacological Evaluation of Novel Antimalarial Targets 
Organisation GlaxoSmithKline (GSK)
Country Global 
Sector Private 
PI Contribution Joint research project under the GSK-Crick Partnership - Pharmacological Evaluation of Novel Antimalarial Targets
Collaborator Contribution Joint research project under the GSK-Crick Partnership - Pharmacological Evaluation of Novel Antimalarial Targets
Impact Half time GSK chemist on the project
Start Year 2016
 
Description Kathy Niakan - Developing an embryo-centric minimal medium for improved human ES and iPS cell culture 
Organisation Cell Therapy Catapult
Country United Kingdom 
Sector Charity/Non Profit 
PI Contribution Research project - Developing an embryo-centric minimal medium for improved human ES and iPS cell culture
Collaborator Contribution Research project - Developing an embryo-centric minimal medium for improved human ES and iPS cell culture - establishing ways to develop the cell culture medium for commercialisation.
Impact Collaboration continues.
Start Year 2017
 
Description Kathy Niakan - Developing an embryo-centric minimal medium for improved human ES and iPS cell culture 
Organisation Stemcell Technologies
Country Canada 
Sector Private 
PI Contribution Research project - Developing an embryo-centric minimal medium for improved human ES and iPS cell culture
Collaborator Contribution Research project - Developing an embryo-centric minimal medium for improved human ES and iPS cell culture - establishing ways to develop the cell culture medium for commercialisation.
Impact Collaboration continues.
Start Year 2017
 
Description Lucy Collinson - MiniLM: A new tool for locating cells expressing fluorescent proteins for ultrastructural analysis by Correlative Light and Electron Microscopy (CLEM) 
Organisation Boeckeler Instruments
Country United States 
Sector Private 
PI Contribution Development of a new tool for locating cells expressing fluorescent proteins for ultrastructural analysis.
Collaborator Contribution Expertise and support around developing the technology for commercial use.
Impact RMC Boekler have added the device to their portfolio as an integrated add-on for their microtomes. This product is now on the market.
Start Year 2017
 
Description Lucy Collinson - MiniLM: A new tool for locating cells expressing fluorescent proteins for ultrastructural analysis by Correlative Light and Electron Microscopy (CLEM) 
Organisation Carl Zeiss AG
Country Germany 
Sector Private 
PI Contribution Development of a new tool for locating cells expressing fluorescent proteins for ultrastructural analysis.
Collaborator Contribution Expertise and support around developing the technology for commercial use.
Impact RMC Boekler have added the device to their portfolio as an integrated add-on for their microtomes. This product is now on the market.
Start Year 2017
 
Description Lucy Collinson - MiniLM: A new tool for locating cells expressing fluorescent proteins for ultrastructural analysis by Correlative Light and Electron Microscopy (CLEM) 
Organisation Kingsview Optical
Country United Kingdom 
Sector Private 
PI Contribution Development of a new tool for locating cells expressing fluorescent proteins for ultrastructural analysis.
Collaborator Contribution Expertise and support around developing the technology for commercial use.
Impact RMC Boekler have added the device to their portfolio as an integrated add-on for their microtomes. This product is now on the market.
Start Year 2017
 
Description Lucy Collinson - Next generation automated ultramicrotomy for high throughput array tomography 
Organisation Delmic Ltd
Country Netherlands 
Sector Private 
PI Contribution Research project - next generation automated ultramicrotomy for high throughput array tomography
Collaborator Contribution secondment of researcher
Impact Proximity to Discovery funds
Start Year 2017
 
Description Maximiliano Gutierrez - GSK partnership - Effect of pharmacological manipulation of target of tuberculosis 
Organisation GlaxoSmithKline (GSK)
Country Global 
Sector Private 
PI Contribution Joint research on the effect of pharmacological manipulation of target for tuberculosis.
Collaborator Contribution Effect of pharmacological manipulation of target for tuberculosis - GSK provision of expertise, technologies and access to libraries
Impact Continued collaboration with GSK
Start Year 2017
 
Description Paola Scaffidi - collaboration with AZ - Modulation of histone H1.0 levels to induce cancer stem cell differentiation 
Organisation AstraZeneca
Country United Kingdom 
Sector Private 
PI Contribution Joint research on Modulation of histone H1.0 levels to induce cancer stem cell differentiation.
Collaborator Contribution Joint research on Modulation of histone H1.0 levels to induce cancer stem cell differentiation. Provision of expertise, technologies and libraries.
Impact In kind support from AZ in form of part time research scientist
Start Year 2017
 
Description Peter Parker collaboration with AZ - target validation - a proof of concept screen 
Organisation AstraZeneca
Country United Kingdom 
Sector Private 
PI Contribution Joint research project - target validation - a proof of concept screen
Collaborator Contribution Joint research project - target validation - a proof of concept screen. Provision of expertise, technologies and libraries.
Impact Further funding from AZ (£75k)
Start Year 2017
 
Description Predicting personalized glucose responses for the diagnosis of metabolic syndrome and Type II diabetes by artificial intelligence 
Organisation University of Cambridge
Department MRC Epidemiology Unit
Country United Kingdom 
Sector Academic/University 
PI Contribution We have developed a mass spectrometry-based MS/MSall approach that on the basis artificial intelligence can extract vast amounts of molecular information from complex biological samples. The extracted molecular signatures are used to train machine learning algorithms and to associate them in this way to phenotypic information. Briefly, we use two-step approach, first we use deep convolutional neural networks to train a machine learning predictor (over a million of CPU hours spend) enabling identification of systemic noise coming from particular biological information, including noise that is associated with particular type of mass spectrometer, and then subtract a combination of identified noise functions from the unobserved sample. What remains is the biological signal that we use further to train a next machine learning classifier that is then trained to predict the phenotypic outcome. Using this approach we recently managed for the first time to predict metabolite concentrations from enzyme abundance upon the sequential deletion of all protein kinases yielding, for the first time quantitative prediction of genotype-phenotype relationship out of a genome-spanning phenotype in Saccharomyces cells (Zelezniak et al, currently under review at Science). We apply for this i2i to evaluate the technical feasibility of our technology towards developing a viable business case with enough evidence to be competitive for Innovate UK funding towards creating a simple diagnostic test for the probability of developing Type II diabetes. Already we have performed market research with MBA graduates at Cambridge Judge Business school, where the aim was to identify feasible market segments for our technology. We identified several feasible business models that could be potentially be viable in the case if technological feasibility of technology that we aim to address here. Further, we are in contact with nutritional counsellors at Igennus (www.igennus.com) to identify potential routes to market also in the lifestyle sector, and constantly talking to numerous patients with diagnosed diabetes to understand their needs to elaborate the best market placement.
Collaborator Contribution Our clinical partner is the MRC Epidemiology Unit, represented by Dr. Langenberg, MRC Program Leader, and Prof. Wareham, who is the co-Director of the Institute of Metabolic Science in Cambridge. Our collaboration partners are internationally recognized experts on the aetiology and prevention of obesity and diabetes, and very excited about the technology and looking forward to collaborate in the study by providing relevant samples and helping interpreting results in statistical epidemiology and clinical pathology of diabetes (see support letter). The Wareham laboratory will support us in the statistical and epidemiological interpretation of the results obtained. It will also provide data to link the molecular signatures identified to established genetic and epidemiological markers.
Impact N/A
Start Year 2018
 
Title Establishment of a MRI-based approach to evaluate the bone marrow vascular functionality as prognostic biomarker in acute myeloid leukemia 
Description Background: In our laboratory a thorough characterization of the BM vasculature in AML PDX using intravital two-photon microscopy has revealed a severe functional toxicity. Vascular wall barriers responsible of oxygen, nutrients and drug delivery appear severely damaged, with increased permeability and hypoxia, altered perfusion and release of normal HSCs to periphery. At molecular level endothelial cells signature is altered, resulting in increased nitric oxide (NO). Moreover, we found persistent vascular leakiness and increased NO levels in AML xenografts after induction therapy in PDX, suggesting that the abnormal and poorly functional vessels participate to maintain a hypoxic vascular environment during the remission phase, likely contributing to treatment failure. Our hypothesis is strengthened by the analysis of NO level in a cohort of patient-derived BM biopsies, which faithfully represents the high diversity of the AML genetic landscape. Our analysis argues for NO as a potential BM biomarker, as increased levels of NO post-induction therapy is associated with a poor clinical outcome. Combination in preclinical models of NO inhibitors to restore normal vasculature and chemotherapy to target leukemic cells successfully delayed disease development and increase survival in PDX (Passaro et al., Cancer Cell in press). Aim and preliminary data: Our results suggest that the detection of a pathologic vascular phenotype in the BM of patients would be of high clinical value as a prognostic factor in AML to evaluate the clinical outcome and direct for the appropriate treatment. Two-photon microscopy being an imaging approach not translatable to humans, we identified magnetic resonance imaging (MRI) as a non-invasive imaging approach to visualize and quantify vascular function in patients. MRI is a medical imaging technique to visualize the anatomy and the physiology of the body using applied magnetic fields and the intrinsic properties of atomic nuclei of certain elements contained in the body (e.g. hydrogen in water and fat molecules). It represents a powerful diagnostic and prognostic imaging tool for many diseases, particularly brain affecting disorders, cardiovascular, liver and gastrointestinal pathologies, and cancer. In particular, it has been suggested as a valuable tool for evaluating the vascular function in solid tumors and is currently tested in many clinical trials (Jain R. et al., NMR Biomed. 2013; Johnson et al., British Journal of Cancer 2016). MRI imaging of the BM has been mainly used to assess the cellularity and the compositional differences between red and yellow marrow, fluctuating over time depending on age and homeostatic requirements. Our aim is to use MRI to assess BM vascular functionality, in particular vascular permeability, perfusion, blood flow and oxygenation. The project will be performed at the Francis Crick institute by Dr Diana Passaro (post-doc in Dr Dominique Bonnet laboratory) and Dr Bernard Siow (Head of MRI at the In Vivo Imaging facility). The project will include a first phase of development and optimization of MRI techniques and a second phase of comparative analysis of healthy animals vs AML PDX vs UCB-PDX before and after chemotherapy. The development and optimization phase has already produced preliminary results supporting the feasibility of the proposed approach. We could obtain high resolution anatomical images of the BM of healthy NSG mice, quantitatively map functional parameters and perform dynamic contrast-enhanced (DCE) imaging using gadolinium as contrast agent. 
Type Management of Diseases and Conditions
Current Stage Of Development Initial development
Year Development Stage Completed 2018
Development Status Under active development/distribution
Impact Acute Myeloid Leukemia (AML) is the most common acute leukemia in adults. While the clinical presentation is quite uniform, it is a highly heterogeneous disease at the genetic level. There is a consistent effort aimed to characterize each genetic subgroup and design the best therapeutic strategy. However, the common clinical practice remains an induction therapy with cytarabine (AraC), still associated to high incidence of therapy resistance and relapse. Our laboratory has recently shown that the clinical behaviour of AML is influenced by the interaction with an altered vascular microenvironment in the bone marrow (BM), which represents an intriguing source of potential therapeutic targets. Using intravital two-photon microscopy, we showed that AML patient-derived samples belonging to different genetic subgroups induced a common pathologic vascular phenotype in the BM of xenografts (PDX). We identified increase nitric oxide (NO) as one of the molecular mechanism associated with vascular toxicity in AML. Moreover, induction chemotherapy failing to restore normal NO was associated with a poor prognosis. Restoring normal vascular functionality in the BM of AML-PDX with nitric oxide inhibitors (NOS) preserved normal stem cell function and improved treatment response (Passaro et al., Cancer Cell in press). Our data suggest that the detection of this pathologic vascular phenotype in the BM of patients would be of high clinical value for AML as a biomarker helping to predict the clinical outcome and direct the appropriate treatment. The sole measurement of NO in the BM might not completely represent the various aspects of the vascular toxicity in all AML patients, more likely to be a multifactorial pathology. Moreover, dosage of biomarkers in biopsies is often reductive and not representative of the whole organ. Vascular function can be imaged in patients with non-invasive magnetic resonance imaging (MRI), a powerful diagnostic and prognostic tool for many diseases. MRI in the human BM is a feasible approach, however it is mostly used to detect changes in the cellularity, fat content and fibrosis. In small rodent, MRI has mostly been used in rats, with only few attempts in mice. Our aim is to set up conditions for using MRI as imaging approach to visualize the AML-derived pathologic vascular phenotype in the BM of PDX, using AML patient-derived samples provided by our clinician collaborator Prof John Gribben (St Bartholomew's Hospital, London, UK). The achievement of this goal together with our previous work (Passaro et al., Cancer Cell in press) will provide strong pre-clinical evidences to start a clinical trial including vascular functionality as an important biomarker to evaluate in AML, and NOS inhibitors in combination with chemotherapy to achieve full remission in patients 
 
Title Insertion-and-deletion-derived tumour-specific neoantigens as predictive biomarkers and novel therapeutic targets 
Description Regarding the CPI predictive test technology, our group has demonstrated fs-indels to be significantly associated with CPI response across three separate melanoma cohorts (P=4·7 × 10-4), and found to be a superior predictor over and above currently accepted biomarkers such as non-synonymous single nucleotide variant (ns-SNV) count (ns-SNV P=4·8 × 10-3 in same datasets). The immunogenicity of fs-indels is believed to stem from the fact they trigger novel open reading frames and a large quantity of mutagenic peptides highly distinct from self. Indeed, analysis of tumour-specific neoantigens (which are hypothesised to be a key antigenic stimulus for CPI activity) showed that fs-indels created three-fold more high-affinity neoantigen binders as compared to ns-SNVs. Furthermore, neoantigens derived from fs-indels are nine-fold enriched for mutant specific binding (i.e. mutant but not wild-type allele predicted to bind to the MHC). In addition, we have also demonstrated an association between fs-indel load and overall survival in non-small-cell lung cancer (high vs low fs-indel load: HR 0·25; P=0·045). Finally, we have shown fs-indel neoantigen presence to be associated with upregulated expression of antigen presentation genes in renal cell carcinoma (RCC), which correlated (r=0·78) with T-cell activation as measured by CD8-positive expression. Regarding the fs-indel therapeutic technology, we conducted a pan-cancer analysis of TCGA data and found >500 cases (>10% of all tumours) to have high affinity fs-indel neoantigens across a narrow list of just 15 tumour suppressor genes. Targeting fs-indel neoantigens via adoptive T cell or peptide vaccine strategies may represent an attractive and broadly applicable therapeutic option. Furthermore, by virtue of being founder events, many alterations in tumour-suppressor genes are clonal, present in all cancer cells, rendering them compelling targets for immunotherapy. 
Type Diagnostic Tool - Non-Imaging
Current Stage Of Development Initial development
Year Development Stage Completed 2018
Development Status Under active development/distribution
Impact Despite the recent clinical success of immunotherapy in multiple tumour types, through strategies such as immune checkpoint inhibition, only a subset of patients currently achieve a durable clinical response. With checkpoint inhibitors (CPIs) costing £100,000 or more per patient, improved patient stratification is required to ensure only patients who are most likely respond to CPIs receive the therapy, providing economic benefit to healthcare systems and sparing non-responders from the unpredictable and potentially lethal toxicities. In addition, expanding the fraction of immunotherapy responders to a much wider patient population is also a current challenge in the immune-oncology field. Addressing these two challenges is contingent upon i) the identification and validation of robust biomarkers predictive of patient response, and ii) the discovery and development of novel immunotherapeutics. The Crick has recently developed two technologies based on recent research findings arising from work performed in Charles Swanton's laboratory (Renal TRACERx) centred on the role fs-insertion/deletion mutations (fs-indels) in tumour immunogenicity and CPI response (1). The first technology represents a predictive test/biomarker that uses the number (or fraction) of frameshift insertion/deletion mutations (fs-indels) within a tumour to predict response rates to CPIs such as ipilimumab, nivolumab and pembrolizumab, wherein a high number fs-indels compared to a reference sample indicates increased chance of drug response. The second technology refers to a therapeutic method that targets fs-indel derived neoantigens. This technology is based on the same datasets as above, which imply targeting fs-indel neoantigens specifically as opposed to single-nucleotide variant (SNV) derived neoantigens will give rise to superior efficacy. The focus of this i2i project will be to produce data that can further exemplify and improve both technologies, which have been patented, by improving and refining our understanding of fs-indels as biomarkers and as neoantigen targets, resulting in a more reliable predictive test and enhanced immunotherapeutic. This will be achieved by investigating the role of nonsense mediated decay (NMD) and the utility of long read sequencing technologies in fs-indel detection. Furthermore, a comprehensive multivariate statistical model will be developed from the arsing results, which will form the basis of a commercialisable predictive test. The rationale to develop these technologies is to provide a solution to the two aforementioned clinical unmet needs. The fs-indel predictive test will enable improved CPI patient stratification and hence identify the patient populations that would benefit from CPI treatment leading to both patient and economic benefits. In addition, the fs-indel based therapeutic method may provide a new innovative targeted neoantigen therapy possessing improved therapeutic efficacy compared to current and competing technologies and hence expand the population of patients who will respond to immunotherapeutic intervention. 
 
Title HERON - artificial intelligence for scientific data mining - Facilitating accessible interdisciplinarity 
Description Accessing and comparing quantitative experimental details form a daily expanding literature is becoming unachievable. In science, the ability to have a general understanding of what happens is paramount. Rapid progress requires awareness or at least immediate access to what has already been tested and discussed. Today, scientific publications are the communication media between researchers and represent the reference point against which we judge what is known, what the trends are and where our focus should lie. Today, scientists lack the necessary tools to make sense of the vast amount of research produced globally. The situation becomes even more severe when overwhelmed researchers navigate the ever-growing stack of scientific literature in search for relevant insights towards their specific scientific questions. To cope, researchers will ignore 99% of the publications based on personally developed exclusion algorithms which might or might not work and will directly contribute to the dangerous problem of reproducibility in life sciences. Also, with the increasing adoption of open-access publication schemes, there is a pressing need for a robust and scalable framework which allows both experimentalists and theoreticians to identify key parameters in scientific data. The project presented here aims to build up a scientific corpus of experimental words and parameters by extracting them from 1000 open-source publications, in the field of neuroscience. In the present approach, typical journal articles are two-part constructs: one containing the interpretations made by the authors based on reported experimental results and the second representing the actual experimental conditions composed largely of biological, chemical and physical constants. Output from prototype: Reported experimental conditions include several major techniques used in the study and their respective operational parameters. Therefore, a tree-like structure efficiently renders their relationship, belonging and separation respectively. Such tree structures can be joined to a graph database (e.g. Neo4J), which provides a straightforward way to generate recommendations based on commonality parameters, make comparisons and research the general global structure of the available data. To test the utility of the concept of a graph based user interface, we extracted data from the entire olfactory literature in the open PubMed database (7280 papers, 2480 keywords). Using a custom written Java software, we converted the XML database into a Neo4J graph, and checked the results on the built-in graph rendering engine of Neo4J. Feedback from test users showed, that although this graph only contained a small number of keywords, it accelerated the process of getting a good overview of the literature on a given topic. Experimental values described mostly in the Materials and Methods sections are composed of numeric values containing the operational parameters and several word constructs defining their meaning and relationship. These parameters are not contained in currently available databases and therefore need to be extracted from the text using semimanual or automated nature language processing (NLP) and AI tools. We have tested the feasibility of this method using OpenNLP (Apache) on 10 neuroscience related papers, and found that experimental parameters can be extracted with good reliability. Project approach: Here we aim, with the aid of human curators and document annotation tools to extract all the experimental conditions e.g. animal model, genotype, cell type, temperature, concentrations, injected current, impedances, etc. from individual publications, to facilitate building of an AI-driven document parsing engine. Curators using publicly available PDF annotation tools will highlight and extract experimental data into a 2-dimensional graph comprising numerical and string values describing i.e. the patch clamping protocol in an individual publication. Each paper is represented by a database of the reported experimental parameters and protocols. Our goal will be, based on the generated annotated corpus, to create automatic methods for data extraction by utilising the quickly developing field of scientific text mining for tagging and parsing experimental sections in publications. Available tools include Geneview for biological entity recognition, AnatomyTagger for anatomical information tagging; NeMine for gene and protein name recognition; ChemicalTagger an open-source tool that uses Open Source Chemistry Analysis Routines (OSCAR4), which will be combined with natural language processing tools and machine learning to extract all relevant data. 
Type Of Technology Software 
Year Produced 2018 
Impact Accessing and comparing quantitative experimental details form a daily expanding literature is becoming unachievable. In science, the ability to have a general understanding of what happens is paramount. Rapid progress requires awareness or at least immediate access to what has already been tested and discussed. Today, scientific publications are the communication media between researchers and represent the reference point against which we judge what is known, what the trends are and where our focus should lie. However, in 2017 at the current rate of 2.5 million papers per year, the publication system is collapsing under its volume. With the ever-wider adoption of open access publications, a record number of predatory or fake journals producing high volumes of poor-quality research and the pressure of "publish and perish" generate a jungle of papers filled with buzzwords for the scientist to wade through on a daily basis, most of which remain unread. Today, scientists lack the necessary tools to make sense of the vast amount of research produced globally. Currently, a group leader at the Francis Crick Institute, to stay completely up to date on his/her field of biomedical research should read on average in excess of 150+ papers per day. In practice, this number will vary with research subject and can be reduced by search filtering, exclusion by journal impact factor, author and lab preference, and further personal bias. However, a mere 10% of the above number already consumes most of one's available daily working hours. With an average reading speed of 228±30 words per minute and excluding matters of comprehension, reading 15 papers containing 6000 words (Nature) would require 6 hours of continuous reading on every day of a calendar year. Solution: To accelerate scientific progress, we propose to further develop Heron, an early prototype AI-driven search engine. In contrast to current list-based practices, Heron features a graph-based user interface where information is stored in the form of nodes and in the multidimensional relationships between these nodes. Nodes are populated with experimental details reported by authors under the Materials and Methods section of a publication and generally contain the list of chemical, biological and physical materials and methods necessary for the complete reproduction of the said finding. Heron's innovative approach enables unprecedented literature-scale examination of information, thus, promoting experimental standardisation and initiate progress on biomedical reproducibility issues particularly afflicting interdisciplinary fields such as neuroscience. Scientific case: Heron, unlike traditional keyword-based online repositories such as PubMed, aims to give an instant structured overview of individual and related publications based on customizable commonality parameters such as anatomical region of interest, genotype, instrumentation, protocols, etc. thus, drastically reduce the retrieval time of relevant information. Heron's user-interface visually aids browsing by displaying papers as nodes in a two-dimensional map. It recommends studies employing similar experiments and allows easy, direct comparison of experimental details. It provides the users with tools to give feedback about the data effortlessly, and finally, allows scientists to add new, unpublished data into the database. Given its scalable and dynamic web-based interface, Heron offers real-time, unrestricted access to scientific information, with two primary functions: Firstly, it is an advanced literature browsing tool. Secondly, it is a disruptive scientific publication tool replacing conventional text-based communications with a transparent and robustly parameterized experimental graph-structure. Natural language processing tools capable of extracting data from scientific text are available and rapidly progressing. Current projects, however, are mostly aimed at a narrow field and limited set of parameters. Heron's primary goal is to achieve a comprehensive and interdisciplinary database of published and unpublished scientific experiments, by combining field expertise and machine learning tools, and pave the way to a universally available, open scientific communication model. Commercialization: The product will be of interest for various user groups and commercial partners. Subscribers will include independent researchers, who can plan their projects with more insight into literature, and assess the feasibility of their technical approach. PhD students and postdocs can get immediate feedback on their experiments, and identify critical parameters when trying to reproduce published results. Medical professionals and scientific journalists can keep up with the literature with minimal effort, and get insight into the data behind protocols and recommendations. User search activity provides the source for highly tuned targeted advertising. Finally, scientific funders can get detailed insight into trends from our graph based data. 
 
Title jULIEs - juxtaneuronal Ultra-Low Impedance Electrodes 
Description jULIEs are composed of ultramicroelectrodes (UMEs) that are <25µm in overall diameter and can transmit current through a solid metal core. Initially for use as a neural probe but has potential for broader use. 
Type Of Technology New/Improved Technique/Technology 
Year Produced 2017 
Impact Prototypes have been developed and are being distributed to research labs. 
 
Company Name METACOGNIS LIMITED 
Description Artificial Intelligence (AI) for Scientific Data mining - Facilitating accessible interdisciplinarity 
Year Established 2016 
Impact Accepted onto KQ Labs Accelerator as a spin out (Scheme funded by Innovate UK) and awarded £40k.
 
Company Name ERVAXX LIMITED 
Description Will leverage novel insights into the expression of human endogenous retroviruses in different cancers to develop a pipeline of first-in-class cancer vaccines. 
Year Established 2016 
Impact Raised £12m