How a Fortune pharma giant successfully used AI & machine learning models and NLP to remarkably shorten drug development cycle, improved the First In Human (FIH) decision making process, and scale-up cross-functional collaboration to effectively navigate the COVID-19 crisis
Drug development has historically been a very cumbersome process, in no small part due to the human casualties that serve as the price for even the smallest of mistakes. Although clinical trials form the basis for this selection, it is preceded by an equally important process referred to as the ‘First In Human’ phase.
After identifying and isolating the relevant compound, the First In Human phase marks the first test for the experimental drug’s effectiveness.
This phase is plagued with a slew of issues ranging from collaboration inefficiencies to non-standardized decision making, all of which contribute to a longer development cycle. Thankfully, we live in an era where technology can be used to iron out these inefficiencies and drastically decrease the time-to-market of potentially life-saving compounds.
The Challenge
Collaboration and Reconciliation Issues
The setting up of the First In Human (FIH) committee introduced several new variables into the drug evaluation process. It results in a concomitant increase in issues relating to collaboration and alignment. This inevitably translated to unprecedented increases in the time taken to achieve a consensus.
Non-Standardized Decision Making
Until the FIH phase, there was no guarantee that all preceding trials would have adopted standardized decision-making tools. Orienting these findings across a singular dimension of interpretation took away a good chunk of time and effort on the part of the committee before it could have even evaluated the said results.
Lengthy Compound Development Cycles
Analog methods adopted earlier means that it’s challenging to reconcile recent findings with prior trials and preclinical data. This, along with the inability to view toxicity connection points between the numerous therapeutic areas, resulted in a much longer development cycle.
The Solution
Computer-Aided Standardization
Our solution began with the clinical data collection process and introduced standardized data intake mechanisms via smart forms. This data was then mined altogether to collate data across the multiple subsections of pharmacovigilance: non-clinical pharmacology, toxicology, DMPK, and clinical pharmacology.
Platform Approach to Garner Insights
Compelling insights can only be garnered by integrating findings across toxicology, clinical pharmacology, and preclinical safety pharmacology. To enable this, we built a custom platform that allowed researchers to employ dynamic, interactive, and intuitive visualizations, which allowed for more nuanced approaches in interpreting said data and using it for FIH decisions.
The Impact
The platform that we had designed for our client helped them leapfrog their way through the mundane and time-consuming processes hindering the successful and timely completion of the FIH process. The committee was no longer burdened by collaboration and alignment problems. The researchers had access to new and sophisticated ways to extract meaningful insights from vast data sourced from the preclinical trials.
Moreover, the platform’s effectiveness lied in the combined effect of its simple, intuitive user interface and the advanced capabilities that it brought to the researcher’s arsenal. Some of the most noticeable results that the platform has had are detailed below.
Advanced Filtering
Starting with study types, researchers were now able to re-order and filter the study types in any order they deem fit. The concentration of the assays within specific study types could now be plotted against linear and logarithmic ranges. The platform also allowed researchers to apply free fractions to determine the level that works best.
Optimize Drug Risk-Benefit
The platform could calculate exposure to effectively determine the dose that strikes the right balance between drug efficacy and adverse effects. Its capabilities extended to calculating the exposure multiples across two studies and then comparing them with each other.
Versatile Visuals
With so many filters, values, and findings to juggle around, there was a pressing need for a sufficiently versatile data viewing and slicing feature that researchers could make use of on the fly. Our platform allowed for different views to categorize across human, species, and non-species studies and boasts of tag and color coding to bestow researchers with the ability to define their filtering framework for convenience.
The platform also came with an auto-translation of nM to ng/ml, the ability to toggle between total and free-form values. This pinned tooltip feature provided information about points and ensures that IB documents are tagged to the corresponding assays.
Other Case Studies
