Automation and standardisation for your qpcr testing

Transcription

1 Automation and standardisation for your qpcr testing

2 Using Artificial Intelligence (AI) to make qpcr fully-automated and safer pcr.ai ensures that your laboratory, test and machine are at the cutting-edge of diagnostic technology. Our ground-breaking patented technology enables AI-based automation of diagnostic test results; making tests more accurate, standardised and safe as well as quicker and cheaper. pcr.ai has been validated as having above 99.9% interpretation accuracy in multiple independent studies with over 0.25 million samples. With pcr.ai you can stay ahead of the curve and bring your lab into the future of diagnostic technology. 3

3 pcr.ai eliminates risks labs and manufacturers face due to the current need for manual analysis Most qpcr test results are analysed by people but laboratories must ensure results are error-free and standardised. A complex chain of quality control (QC) is therefore required, in the form of specialists checking on specialists. This dependency on a chain of highly-trained staff, monitored by clinicians, is expensive, reduces availability to standard working hours, makes it difficult to implement new tests and increases turnaround time for samples. The need for manual analysis increases the risk of human error. This is contributory to diagnostic error rates, which can be as high as one in five. 1 For labs and manufacturers, diagnostic errors pose an unacceptable commercial risk, as patient misdiagnoses can lead to suboptimal treatment plans and in the worst cases, preventable death. An error can mean loss of business, noncompliance with regulatory requirements, laboratory closures 2 and in some cases even injury to patients (and resultant lawsuits). Manual review of manual analysis is the current status quo To mitigate these risks, clinical laboratories invest heavily in data-analysis and software training, QC and staff evaluations. 3 Manufacturers invest in creating complex guidelines for each test and platform. Without an automated solution, labs have little choice but to continue with expensive and risky manual analysis. Of the ~3bn qpcr tests performed every year, regardless of machine, the majority are analysed manually. pcr.ai automates analysis of qpcr pcr.ai eliminates the risks and costs that manual analyses pose to labs through innovative use of AI. pcr.ai can increase diagnostic accuracy, cuts costs and save time by providing AI-based automation and standardisation of qpcr results for all tests and machines. 1 F.J. Hamilton et al; A comparison of the Vela Sentosa Quantitative CMV assay with an in-house quantitative CMV assay; Journal of Clinical Virology, Volume 70, Supplement 1 (Sept. 2015). 2 For example, a large Canadian hospital JGH, terminated a $450,000 MRSA test contract with BD due to correct positive rates of only 40%, the CEO at Theranos is not currently able to run a blood testing lab (and the company has closed all their blood testing labs), and so on. 3 M. J. Espy et al; Real-Time PCR in Clinical Microbiology: Applications for Routine Laboratory Testing; Clinical microbiology reviews (Jan. 2006). 4

4 pcr.ai reduces risks, costs and analysis time Reducing risks pcr.ai reduces the risk of a diagnostic error occurring. It provides interpretation accuracy of 99.9%+, as validated in multiple independent studies. AI-based automation of test analysis makes tests safer by lowering staff involvement, which reduces the opportunity for human error and variability. It also enables senior staff to delegate more work to juniors with knowledge that issues will be detected before they become problems. pcr.ai makes tests more standardised, accurate and safe, Lowering costs AI-based automation of results analysis makes running qpcr tests simpler and quicker. In today s resource constrained environments, this gives junior staff more autonomy to manage the full workflow and frees more senior staff to focus on the many other tasks they have. A recent cost analysis with a large NHS hospital showed that pcr.ai could save more than 200,000 per year in staff costs, especially given the new regulatory environment that requires ISO15189 adherence (that is becoming law in the UK in 2018). Saving time Independent studies 4 have shown that pcr.ai saves an average of 60 minutes of specialist hands-on time, with result interpretation taking fractions of a second and analysis available 24/7. Unrestricted by regular working hours, pcr.ai makes testing speedier and this allows labs to perform more tests. This is critical for labs that are increasingly pressurised by ever-growing workloads and challenging deadlines. pcr.ai benefits pcr.ai helps improve patient safety by making diagnostic results more automated, standardised, quicker and cheaper- this helps improve patient safety. Key benefits include: Improved accuracy and standardisation for lab-developed tests and commercial kits. Time savings- pcr.ai is ten times quicker at interpreting results than experienced professionals doing this manually. Savings of 90% of hands on time with 24/7 availability. Lowered lab costs by decreasing the dependency on expensive specialists who are in short supply globally. Efficient spotting and troubleshooting of processing and material testing errors due to built-in monitoring (i.e. automated Westgard, etc.). Makes current controls more effective due to automation and standardisation. Easier maintenance of largescale testing standards and patient movement due to high results reliability. Effective adherence to legal regulations enabled by fullautomation of tests, especially stricter EU (ISO15189) and FDA standards that are being introduced soon. 4 RG Gunson et al AccuCall : A Novel Solution for the Automated Interpretation and QC of in-house, Real Time PCR Results; Journal of Clinical Virology, Volume 70, Supplement 1 (Sept 2015). 5

5 pcr.ai makes existing controls more effective Labs are pressurized environments with limited time and staff to scrutinize runs. Using known sample control materials (molecular controls) helps to monitor and minimise incidences of problems including material, processing and machine issues. However, these can be overlooked or misinterpreted, leading to misdiagnoses or need for repeats. pcr.ai provides immediate alerts when controls are not performing as expected. This enables labs to quickly and efficiently spot and troubleshoot problems before they negatively impact results. Examples of rare but expensive issues that have been detected in a timely manner include machine issues such as obscured sensors. pcr.ai is a new type of AI-based software control that ensures that existing molecular control results are continually monitored in real-time on a machine-by-machine, test-by-test and unlimited timespan basis. This step change in quality control eases adherence to evolving government regulations. Variable lab results from the same sample are common Commercial kits are often not fully automated There are kits available for clinical usage for a limited number of pathogens. Kits may be more expensive or less flexible than inhouse developed tests. However, many of these automated kits do not include data-analysis automation. A few of many examples of commercial assays that require manual input as part of their published operating procedures include the CDC Influenza kit, Focus Diagnostics ASRs and kits, and most others. Using commercial kits does not release the laboratory from the responsibility of ensuring accurate data-analysis. The below analysis by the British government biological standards agency (NIBSC) shows the variability in results for thirty clinical labs that were given the same Norovirus sample. The data shows that two of the laboratories repeatedly failed to detect the virus and a further four had standard deviations of results well beyond those expected from the group. Presented by N Almond, NIBSC at SOGAT

6 pcr.ai self-learns, accurately automating any test AI learning machines are algorithms that enable computers to learn and develop through exposure to new data. pcr.ai works by using an initial calibration stage, guided by answers or results from the test developer or senior laboratory scientist, to sort machine outputs into negative, positive and indeterminate clusters. Based on these results, pcr.ai creates a new analysis template, and uses this to classify new data automatically. This ensures that the same sample for the same test always gets the same expert result, uninfluenced by human variability, ensuring that results are standardised. AI-based automation enables pcr.ai to provide accurate routine analysis in milliseconds, without the need for specialists. Workflow how a test is calibrated Meet minimum data requirements for accurate calibration and appropriate required statistical validation of accuracy*. Assign target and control labels, to ensure that valid templates and controls are used with the assay. pcr.ai automatically sends back initial review and questions on data. Initial calibration performed based on known sample results and client input. pcr.ai compares results to validation data supplied, to ensure required accuracy is reached. Prospective testing performed by client (or partners) to generate any further performance measurements required. *Data can be gathered from either: a. Controlled positive/negative sample dilutions at agreed concentrations (SOP supplied). b. Historic results (for example if the assay has been in clinical use for some time). Workflow - routine use Thermocycler data uploaded or sent automatically to pcr.ai. pcr.ai will analyse and create a report for operator review. Samples marked for rerun can be excluded or manually classified by an authorised member of staff (passwordprotected area based on optional userpermission levels). Export to LIMS if no equivocal samples or reviewer problems remain in the report. Export to Excel and similar formats for further review, anytime. 7

7 pcr.ai how it works (the technical details) Parameter extraction 10 8 A number of parameters are extracted from the curve enabling it to be characterised by a unique fingerprint 5. This is done in two steps as shown in figure A line connecting the first and the last data points of the amplification signal is calculated and the line's values at each cycle are subtracted from the amplification signal. This operation is called a shearing transformation and is used to create a new graph (bottom chart). This new chart allows various parameters such as the start and stop point of the reaction to be more easily extracted. Note that the minimum point on this new curve is the elbow point on the original chart Fig. 1: Curve parameters are extracted Geometric mapping enables curves to be categorised correctly Each curve is mapped onto a new chart (figs. 2 and 3). By calculating the probability of getting from a particular data point to its adjacent points separately via a computer generated random walk, the data can be spread into an enclosed region 6. While a particular positive point may be geometrically near to a particular negative point, the sum of probabilities for any positive point to reach such a negative point would normally be very small, and so these points would be spaced farther apart from one another in the map. This process is known as dimensional reduction and normally decreases the number of parameters which are eventually used in the analysis, as some hardly differ between samples, thus decreasing the computational load for analysis. Fig 2. Curve is represented on multidimensional plot Fig 3: All curves are plotted Routine usage The geometric map can be saved for the analysed assay. A new reaction curve (fig. 4), can be located on the geometric map by applying the random walk method which places it in its most likely area. 5 Fig 4: New curves can be located easily Patent #US , Identifying transition points in chemical reactions Patent #US A1, Automatic chemical assay identification 6 8

8 Performance Details Technical Machines supported Input file types Output file types Data required Open for use with any assay Data storage/backup ABI 7500 (incl HT), ABI 7900 (incl HT), ABI StepOne, ABI Viaa7, BioRad CFX96, BioRad CFX384, Qiagen Rotor-Gene 3000, Qiagen Rotor-Gene 6000, Roche LightCycler2, Roche LightCycler480, Stratagene MX3000/5, Fluidigm BioMark.eds,.sds,.ixo,.rex,.txt,.xlsx,.xls,.csv,.rdml CSV, txt, pdf Raw Fluorescence data Full storage available including logging of user actions for audit. Features Cq Efficiency Absolute quantification Quantification algorithm Relative quantification Graphical display Standard curve * Statistics Quality Control default options Westgard, In-Well validation, Quantification rules, Control validation Performance Run-time Number of users per account 50 curves per second Unlimited Remote access OS Windows XP, 7, 8, 10 Available offline * Regulatory CE-IVD certificated *Upon customer request CE-IVD certified hardware available 9

9 Independent validations Partner Title Publication/Conference Year Results & key findings Johns Hopkins University [Work-in-progress] ,000+ samples processed, comparison with manual shows concordance above 99% NHS Glasgow, UK AccuCall : A Novel Solution for the Automated Interpretation and QC of in-house, Real Time PCR Results RG Gunson et al AccuCall : A Novel Solution for the Automated Interpretation and QC of in-house, Real Time PCR Results; Journal of Clinical Virology, Volume 70, Supplement 1 (Sept 2015) samples; 100% concordance 90% quicker than a technician University of Washington Hospital, USA NHS Glasgow, UK Utilization of Azure PCR AccuCall Software to Improve Analysis of PCR Data Novel Solution Automates qpcr Diagnostics for Qualitative and Quantitative Data without Loss of Accuracy L Hussel et al; Clinical Virology Symposium, USA R Gunson et al; ECCMID, Germany samples, 100% concordance More accurate than auto-threshold automation Quicker and more efficient samples; 100% concordance DuPont Pioneer, USA Novel, Fully Automated Method Allows Efficient Analysis of qpcr Data for Qualitative Calling Based on Comparative Cq Pioneer Hi-Bred Nucleic Acid Analysis Lab; Advances in qpcr, Germany samples; 99.8% concordance St. Georges NHS Trust, UK Automated data interpretation of H1N1/09 ( swine ) influenza Realtime PCR data HPA Pandemic influenza conference, UK ,500+ samples; 99.1% concordance 10

10 pcr.ai product features Standardised results for increased accuracy One test, one result no manual bias from a computer Saves specialist hands-on time Automated interpretation reduces the need for experienced specialists Reduces sample-to-answer time Studies show that interpretation time is cut by 90% Available anytime Run tests overnight and at weekends when specialists are unavailable Available in the cloud (and offline) Secure, encrypted access available anywhere One interface for any common thermocycler and any test Can import run-files from all major thermocyclers for any qpcr test Real-time Westgard reporting and other QC checks Automatically apply all QC checks that currently you do in different systems Unlimited secure storage of results Audit and quality control is simplified with one source for your data Automated and customisable reporting Customise the report to match your laboratory requirements 11

11 pcr.ai and its underlying technology is protected by patents and pending patents. For further information please contact diagnostics.ai. Diagnostics.ai is the trading name of Azure PCR Ltd diagnostics.ai Limited, 59a Brent Street, London, NW4 2EA, W: Company number: