Skip to main content

Artificial Intelligence to transform Drug and Chemical safety testing

Periodic Reporting for period 2 - PATH-TOX (Artificial Intelligence to transform Drug and Chemical safety testing)

Okres sprawozdawczy: 2019-11-01 do 2020-10-31

Every year thousands of new drugs are developed and must undergo preclinical safety studies before clinical trials in man. Significantly, 1 in 4 compounds fail during this important evaluation stage, preventing significant side effects and adverse events in patients. These studies are primarily performed in animals (mostly rodents) due to the complexity of drug/ compound metabolism in humans and are evaluated by highly experienced toxicologist pathologists, who undergo a minimum of 10 years of training. However, there is a skills gap crisis looming due to reducing numbers of experienced pathologists, which is projected to get worse over the next decade, which could delay new drug development.

This project is focused on addressing this major challenge in toxicology pathology, to drive digitisation and efficiencies into the toxicology pathology arena. This PATH-TOX project is led by Janssen R&D and Deciphex Ltd, an Irish SME who developed the technology, in collaboration with PDS, a Swiss-based SME. This PATHTOX project has brought together a multidisciplinary team, essentially to digitise the microscopic analysis of tissues derived from drug safety studies, to drive efficiencies in the system.

In this PATHTOX project, we intended create an innovative pathology Computer-Aided Diagnosis (CADX) system called ‘PATHOLYTIX-PRECLINICAL’ to analyse organs for the presence of toxicity. The system uses Artificial Intelligence (AI) to identify normal/abnormal organs, prioritising abnormal organs for review. Benefits include streamlining workflow to enhance productivity and efficiency for users, saving time for pathologists in whole slide imaging review, slide sharing and expediting drug safety analysis for clinical trials.

At the start of this project, Deciphex had prototype AI technology and required the support of key partners to fast-track the development of the PATHOLYTIX system. Specifically, this project enabled:
- Access to proprietary drug studies samples from Janssen Pharmaceutical, which is vital for the development and validation of AI algorithms.
- Integration with laboratory information management systems (LIMS) which holds treatment/organ/slide information, specifically PDS LIMS platform software called Ascentos and its pathology module Pathdata.
- Accommodate the multi-layered IT security systems in Pharma/CROs, in order to validate the system on-site.

The aim was to design the system based on feedback from pathology users, continuous technology development and validation, in collaboration with partners. This was also be supported by a comprehensive commercialisation, dissemination, and IP plan, to ensure the successful commercialisation of the product.
A streamlined end-to-end workflow for toxcology pathology has been designed and created, namely the PATHOLYTIX PRECLINICAL platform. This workflow was developed from voice-of-the customer analysis and is optimised and refined as per pathologists needs. The accompanying bespoke ergonomic hardware was designed for use with the system. An iterative design process was undertaken to ensure end-user feedback was incorporated in each stage of the design, ensuring inclusion of pathologists’ ideal design features, which in turn sets us apart from current competitors. The system was also designed as a secure cloud system, incorporating several layers of security features, providing protection for data during transit and rest. The impact of this enables capacity to scale the SAAS with large volumes for the delivery of ‘real-time’ analysis of content.

Deciphex has a fully developed AI image analysis module, PATHOLYTIX AI, along with embedded statistical tools and metrics. The platform contains various supervised and unsupervised models, along with multiple data augmentation approaches, that can be used to detect anomalies and toxicities in drug-treated tissues. The team used these tools to develop robust organ recognition and artefact detection. In a large study, the team gathered a significant number of digitised and annotated tissue samples from public and proprietary sources, on which they optimised and validated AI algorithms to detect lesions and toxicities in 5 organs, resulting in a peer review publication in the Toxicology Pathology journal.

Similarly, the system was optimised and integrated with the laboratory information management system (LIMS), facilitated by PDS which was a significant milestone for this project. The LIMS system usually hold all experimental design and results of toxicolgoy studies, hence bi-directional transfer of data is key to full integration into potential customers laboratories. In this project the interactions and data flow between the two systems was mapped and linked. This direct digital correlation to sample metadata streamlines the process significantly and mitigates the need for paper reporting.

At the start of this project, Deciphex’s a priority goal was to implement compliance for ISO 27001 as an Information Security Management System. In November 2019, ISO27001 Certification was achieved by Deciphex and renewed in November 2020. This certification proves Deciphex are able to safeguard customer data and information in the best possible way.

Overall all tasks defined were completed successfully, resulting in a new digital pathology platform built for toxicology pathology.
The primary aim of the project was to develop a toxicology pathology-based digital pathology platform. The novelty of our approach lies in the design, designed by pathologists for pathologists. The extension of the platform to facilitate GLP peer review has also been achieved and is the first of its kind in use. Due to this novel design process and capabilities achieved, the final developed system is beyond the state of the art.

The overall project has made significant impacts in a variety of areas. The project has facilitated the development of innovative technology, supported Janssen R&D pathologists, enabled the fast growth of the Irish SME Deciphex, and demonstrated cost benefits and time-saving capabilities to pharma who have indicated a willingness and commitment to use PATHOLYTIX to support their work.

Concurrently, the impact of the potential of the PATHOLYTIX platform has been commended by Frost and Sullivan, who presented Deciphex with a ‘Global Enabling Technology Leadership Award’. The work has also resulted in a peer-review publication, with more currently being drafted, which will be used to demonstrate the utility of the platform and the AI algorithms developed.

The potential impact of this project has been disseminated at several high-profile pathology gatherings including the European Society of Toxicological Pathology, Global Engage DigiPath Europe, STP 38th Annual Symposium and the 17th European Congress of Toxicologic Pathology. The subsequent engagement and partnership with Deciphex from many large pharma companies and CRO’s has been extremely positive. Strategic partnering and co-marketing opportunities have arisen, increasing brand awareness and trust in Deciphex.

Overall, the project has had a major impact on all partners, the resulting PATHOLYTIX platform allows full integration of AI, LIMS, and GLP compliance, which has the potential to totally transform the toxicology pathology industry by streamlining the entire workflow and supporting pathologists in their pathology reviews.
Anomaly detection using the Patholytix platform