Fractal Therapeutics is building services to make model-based drug discovery and development more effective and efficient. Getting molecules from bench to bedside is a major challenge. Fractal is providing solutions to advance drug discovery by using their model-based analyses, along with cost-efficient animal studies and focused clinical trial designs, to provide scientific input that is rigorous and timely. This involves helping clients answer critical questions posed by the FDA to evaluate therapeutic molecules. Fractal’s services therefore help clients from project inception to clinical proof-of-concept.

Fractal Therapeutics

The COVID wave America doesn’t care about: ‘Everybody is sick of COVID’

Omicron patients may develop Long COVID less frequently than those who had other variants, study finds

12/2020


      10.1371/journal.pone.0250780
    

New immune-evading Omicron variants BA.4, BA.5 are likely in every U.S. state, an expert says. Here’s where they’ve been identified so far

mortality burdens

rapidly waning natural

Patient-derived xenografts (PDX) are an essential preclinical model system in the development of anticancer agents. In our poster from AACR 2024, we developed and computationally validated model-based technique for improving inference about clinical activity based on PDX data.

PDX models have well-validated predictive value for clinical outcomes. Typically, PDX studies are read out in terms of the ratio of treated to control tumor volumes in the interest of predicting clinical Objective Response Rates (ORR). However, in our previous work, we have demonstrated that ORR has a paradoxical relationship with progression-free survival (PFS). While ORR primarily reflects the responsiveness of drug-sensitive subclones within a tumor, PFS is determined by how quickly treatment resistance develops and dominates. As PFS drives drug approval, this is the most important metric to optimize for in preclinical and clinical studies. 
 
This raises the question: Is there a better way to interpret PDX studies to predict PFS, which determines therapeutic potential, rather than ORR? To answer this question, we designed an in silico PDX study and tested methods of predicting PFS. We simulated the PDX study using a tumor kinetics model derived from the literature. With this test dataset, we evaluated the empirical duration of PFS and fitted a tumor clonal kinetic model to derive a model-based PFS. In comparing these experimental outcomes to the true PFS, we found that the empirical PFS tends to over-estimate the true PFS, with over-estimation increasing as sampling intervals increase. On the other hand, the model-derived PFS is significantly more accurate and less biased.   
 
This example illustrates how fit-for-purpose model-based approaches can leverage experimental data more effectively than traditional methods. By fitting a model rather than merely examining the experimental endpoint or raw data, we improve the predictive value of preclinical datasets. If you’d like to learn more about this, feel free to reach out to us on LinkedIn! 
 
#PDX #Translation #DrugDevelopment #Oncology #PFS

How to Keep Schools Open during COVID | SPH - Boston University School of Public Health

medRxiv 2020.11.17.20233726

vaccines alone could not control

“Me too” drug development is derided for being un-innovative and characterized as a slam dunk (often by business school professors with no direct involvement in drug R&D). As it turns out, drug redevelopment projects- biosimilars, generics and reformulation products- are anything but a sure thing, and thoughtful, science-driven development can lead to dramatically improved outcomes. 

While the risk of technical failure for me-too projects is lower, the risk of commercial failure is much higher. For generic and biosimilar drugs, lower prices and a smaller market share can constrain the upside of these drugs, while reformulations must deliver on their technical promise if they are to penetrate the market. So, in a very real sense, me-too drug developers operate within a similar set of strategic constraints as first-in-class companies do. While the smaller expected upside leaves a smaller margin of error, innovative science-driven approaches can be used to reduce the risk of project failure. 

This webinar will dive into ways in which model-based drug development approaches can be used to efficiently leverage reference data from originator compounds in accelerating the path of me-too programs to clinical approval. Topics will include projecting the expected performance of me-too drugs in the clinic, evaluating bioequivalence and therapeutic equivalence efficiently in the clinic, and designing sustained-release formulations rationally. The focus of the talk will be practical, with an emphasis on using the correct strategies to build data packages that can drive rational stage-gate decisions for these programs.

Attendees can expect to come away with a clear understanding of how mathematical modeling, tightly integrated with preclinical and clinical pharmacology, can de-risk and accelerate me-too drug development programs. 

#generics #biosimilars #drugdiscovery #drugdevelopment #webinar #IND #NDA #modelingandsimulation

Cell Reports Medicine 2021 3(3): 100556.

which it did

09/2020

Five years of the COVID-19 pandemic: An interview with Dr. Arijit Chakravarty - WSWS

Most patients enrolled in trials of experimental anticancer therapeutics have late-stage malignancies, which often present as sets of dispersed tumors, each with its own growth kinetics. Detecting and quantifying treatment effect under these complex conditions can be challenging. In particular, assessing a patient’s tumor burden in aggregate over time – rather than evaluating each lesion independently – may result in a loss of information due to inter-lesion heterogeneities. In a recent poster presented at AACR, we explored the potential risk posed by the typical approach of aggregating individual tumor measurements from multiple-lesion clinical data into a single tumor burden measurement.

We used quantitative disease progress modeling to determine how analyzing multiple lesions separately versus in aggregate affects apparent tumor response and the ability to detect progression free survival (PFS) and relapse. We built a two-population clonal dynamics model consisting of sensitive and resistant tumor cells undergoing growth in the presence or absence of treatment. We performed parameter sweeps to assess the impact of individual model parameters on PFS and relapse. We then extended these sweeps to the multi-lesion situation, using nonlinear regression to compare parameter values of aggregated models and the parameter values of the individual models that comprised it.

According to our simulations, the most influential parameters in determining rebound are the growth rate of the resistant cells and the fraction of these cells, while growth rate of sensitive cells and growth rate inhibition affect effectiveness of treatment. In analyzing the multiple lesion model in different situations, we found that estimates for PFS were significantly more sensitive when the lesions were analyzed separately, rather than being aggregated. Our work suggests that aggregating tumor lesions into a single readout neglects the impact of individual lesion growth rates, impairing early detection of resistance as the ability to detect fast growing, small, resistant lesions (despite overall tumor burden decreasing) is lost. Taken together, our findings suggest that relapse can be predicted at earlier times when using the multiple lesion model compared to an aggregated model.
 
#DiseaseProgressModel #Oncology #ClinicalDevelopment

Developing drugs that effectively target the brain is one of the most challenging—and high-stakes—areas of drug discovery. With CNS disorders like Alzheimer’s, glioblastoma, and chronic pain representing huge unmet medical needs, the pressure is on to develop therapies that can cross the blood-brain barrier and deliver meaningful results. But will your drug get there?

Join us on October 24, 2024, at noon EDT for our latest webinar, “Cracking the Code: Model-Based Development of Brain-Penetrating Drugs.” We’ll untangle the secrets of effective CNS drug development, from designing a robust target product profile to leveraging biomarkers for clinical success.

Learn how to navigate the unique challenges of CNS drug development, optimize your screening cascade, and make better go/no-go decisions—all while staying within budget and timelines. This webinar is essential for anyone working in CNS drug development, providing the insights and tools you need to boost your program’s chances of success.

Don’t miss this opportunity to gain a deeper understanding of model-based approaches that can transform your CNS drug development strategy. Register now to secure your spot!

#modelingandsimulation #CNS #drugdevelopment #webinar #brainhealth #Alzheimers #glioblastoma #painmanagement 

An essential consideration in the development of CNS-acting drugs should be the molecule’s extent of blood brain barrier (BBB) penetration, as this determines the drug’s ability to reach its active site. In addition, the BBB may decouple or desynchronize drug kinetics between the body and brain, raising questions regarding the impact of scheduling on CNS active drugs. In this poster presented at AACR, we investigated the relationship between dosing schedule and expected efficacy and toxicity of therapeutics targeting brain tumors.  
 
We developed a two-compartment pharmacokinetic (PK) model to simulate the exchange of a chemotherapeutic between the body and the brain. We coupled this brain PK model with a tumor progression model in which tumor cells exhibit binary resistance to the drug under a logistic growth model. The drug reduces the growth rate of the sensitive population, while the resistant population is unaffected by the drug but suffers a small growth penalty due to the mutation conferring resistance. We predicted regime efficacy based on the tumor progression model and toxicity based on a model of neutropenia. 
 
We found that although the BBB complicates prediction of drug activity, it simplifies schedule effects by damping oscillations in brain concentrations. Generally, drug concentrations in the brain increase with initial treatment until equilibrium between active transport, diffusion, and drug metabolism is reached. Fluctuations in drug concentration resulting from breaks between bolus dosing are greatly reduced in the brain relative to the body, and high frequency dosing—on the order of daily—results in a nearly constant drug concentration in the brain. The rate at which equilibrium is reached and the drug concentration in the brain are unaffected by dosing period as long as the average daily dose is constant. Furthermore, the impact of dosing cycle length on drug efficacy is expected to be negligible.  
 
The limited impact of scheduling on antitumor activity in the brain presents an opportunity to optimize schedule for reduced toxicity. We found that neutropenia is minimized by schedules with dosing cycle lengths that factor evenly into the 18-day neutrophil recovery period. These schedules have consistent toxicity throughout the cycle. Thus, strategies with two, three, six, nine, and eighteen day cycles induce neutropenia least. Although this analysis is based on a hypothetical drug, it provides practical insights that are applicable across a wide range of systemically dosed small-molecule therapeutics. The modeling framework developed here can be readily extended to the guide design of dosing schedules for real-world therapeutic interventions. 
 
For more on this topic, join us for our upcoming webinar on model-based development for brain penetrating drugs! Registration link in the comments. 
 
#AACR #Pharmacokinetics #DrugScheduling #BBB

BA.4 and BA.5, two new Omicron variants sweeping South Africa, detected in U.S.

01/2021

The failed COVID-19 pandemic policies: An interview with Arijit Chakravarty of Fractal Therapeutics - WSWS

medRxiv 2020.12.19.20248544

Surviving the pandemic is only half the battle: ‘Long COVID’ could affect a billion in just a few years

likely to evolve

medRxiv 2020.09.18.20197400

this preprint

Controlling long-term SARS-CoV-2 infections can slow viral evolution and reduce the risk of treatment failure - Nature.com


            
              Public Library of Science
            
          

deeply flawed

Every step in drug discovery brings specific challenges with it, and learning to address these challenges without cutting corners is a key survival skill, one might say. 
 
One such conundrum can be found during lead optimization. During this stage, getting a handle on the pharmacokinetics (PK) of a drug candidate is crucial. PK, which describes what the body does to a drug over time, strongly impacts the utility of drug candidates. The most potent candidate drug in the world will achieve nothing if it doesn’t hang out in the body long enough to do anything! 
 
Getting a clear picture of each candidate molecule’s PK (in fact, its projected PK in humans) is a key piece of during candidate selection. Good-quality PK data can be expensive, however, as it requires a time course in animals (mice, rats and dogs usually). There’s a strong tendency, especially early on in lead optimization, to opt for the “quick and dirty answer”, conducting minimal PK studies with just one or two timepoints to save time and money. However, this approach can lead to misleading conclusions during lead optimization, as we show in the white paper below. 
 
We critically examine PK study designs, highlighting how limited two-point PK studies can fail to identify the most promising compounds. We demonstrate that a simple five-timepoint PK study provides much better estimates of critical parameters like the area under the curve (AUC) and peak concentration (Cmax) compared to an abbreviated two-timepoint design. We also talk about how to use, advanced model-based techniques, such as D-optimal study design, to optimize study efficiency without sacrificing data quality. The number of timepoints, and where to place them, can have a big impact on the quality of PK data that your team generates. In this white paper, we also offer a set of practical design tips to get more out of your PK studies. 
Read the white paper “How to design better PK studies during lead optimization” to learn more about this practically useful skill. This white paper is a teaser for a larger (and complex) topic, that of how to use model-based approaches to squeeze more performance out of your in vivo pharmacology budget. If you’re curious and would like to know more, feel free to drop us a line on LinkedIn! 
 
#Pharmacokinetics #Preclinical #DOptimality

11/2020

would lead to a high disease burden

The U.S. is in a fifth COVID wave, and for many, immunity is waning. Why aren’t second booster vaccines available to all Americans?

preprint

‘Was damage done? I don’t know’: Health experts are slow to criticize Fauci but quick to correct his claim that we are ‘out of the pandemic’

medRxiv 2020.11.05.20226712

medRxiv 2020.08.19.20177550 

medRxiv 2021.03.11.21253375

medRxiv 2021.04.26.21256136


            
              Free PMC article
            
          

Biomarkers are indispensable in clinical drug development, offering a way to predict and assess patient response. However, not all biomarkers are created equal. The utility  of a biomarker in supporting accurate decision-making hinges on its ability to predict response with adequate sensitivity and specificity.  
 
In our latest white paper, we show how Receiver Operating Characteristic (ROC) analysis can be a game-changer in selecting and optimizing biomarkers.  
 
ROC analysis is a statistical method used to evaluate the performance of a biomarker by plotting the trade-off between sensitivity (the true positive rate) and specificity (the true negative rate) across various threshold settings. The area under the ROC curve (AUC) provides a single, interpretable metric that indicates how well a biomarker can distinguish between different patient outcomes. A higher AUC suggests that the biomarker is more effective at categorizing patients accurately. 
 
ROC analysis provides a clear framework for evaluating biomarkers, allowing a  more rational selection of biomarkers and the establishment of meaningful thresholds for clinical use. 
 
This technique can also be used in a PK/PD modeling context, where time-dependent biomarkers can be used to predict patient response across different doses, schedules, and formulations. By applying ROC analysis, we can distinguish between biomarkers with strong predictive power and those with limited utility, ensuring that only the most reliable markers guide our development decisions. 
 
By understanding and applying ROC analysis, drug developers can make more informed choices about which biomarkers to use and how to set thresholds that maximize clinical success. This data-driven approach not only optimizes the use of Phase I data but also improves the likelihood of achieving the desired outcomes in Phase II trials. 
 
If you’re looking to streamline your clinical development efforts and improve your decision-making process with robust biomarker selection, this white paper is a must-read. Click below to read the full white paper and discover actionable insights for your clinical programs. 
 
#ClinicalDevelopment #DecisionScience #Biomarkers #PKPD 

medRxiv 2020.11.29.20240440

The pandemic will never end the way you want it to: ‘It reminds me very much of “The NeverEnding Story”’

medRxiv 2021.04.10.21255251


      PMC8081162
    

better administration of intranasal drugs

critical importance

Adenovirus? COVID? Dogs? What’s behind the outbreak of hepatitis in children?

BA.4 and BA.5, two new Omicron variants sweeping South Africa, detected in U.S. - Fortune

Individually optimal choices can be collectively disastrous in COVID-19 disease control - BMC Public Health

bioRxiv 2022.01.25.477734.

In lean times, companies and project teams face tremendous pressure to move their molecules to the clinic as quickly as possible. There are two widely held beliefs underpinning this strategy: first, that clinical proof-of-concept is critical for sustaining program momentum and second, that speed is of the essence. Unfortunately, only one of these beliefs is correct.

Nine out of ten molecules that enter into the clinic never make it to approval. From the outside, drug development looks like a gambler’s game. After all, if you never know which program is going to succeed, doesn’t it make sense to just put it in patients as quickly as possible and see what happens? A common sentiment is “if the drug is really good, then we should see efficacy signals in the clinic no matter what we do”.

All too often, companies and project teams will fail to focus their effort correctly, over-optimizing preclinical efficacy data for the sake of “creating buzz”, then rush into the clinic only to find that their promising preclinical results fail to translate. Unfortunately, this approach typically ends in disaster for programs and companies.

In a recent article (“Into Thin Air”, link in comment below), we point out why this happens. The TL;DR is that clinical success or failure hinges critically on design choices, such as dose schedule, escalation scheme, and choice of endpoints. When corners are cut during planning for the trial, clinical failure becomes far more likely. That failure can be insidious, and it can often take companies and project teams a long time to realize that they’ve painted themselves into a corner. By the time the consequences of poor planning become obvious, it’s usually too late. (Check out our article “Pitfalls in Clinical Development”, also linked to below, to learn more about this).

There is a better way. Design choices in clinical trials can be made quickly and efficiently using model-based approaches, and are a better career choice in the long run. (Think of it as the Moneyball approach to drug development, rather than the Powerball approach!)

Join us on January 6th, 2025, at noon EST for our next webinar, “Getting to Clinical Proof of Concept”, where we talk about how to design a clinical trial with success in mind.

Learn how to navigate the challenges of selecting the right dose, schedule, escalation scheme, patient population and endpoints in a focused yet scientifically rigorous way. This webinar is essential for anyone working in drug development, from beginner to expert. The insights and tools that we discuss will help you reframe your development strategy to take the risk out of it, while also maximizing “bang for buck” for every dollar spent.

Don’t miss this opportunity to gain a deeper understanding of model-based approaches that can transform your early-clinical development strategy. Register now to secure your spot!
#modelingandsimulation #drugdevelopment #clinicaltrials #webinar #clinicalPoC

“Mathematical modeling doesn’t work. It’s just a bunch of lines fitted to the data points”. If you’ve ever heard someone say this, there’s a good chance that it’s because they’ve seen examples of where modeling got an answer badly wrong. 
  
A famous example of this happened during the early days of the COVID-19 pandemic, when the IHME (Institute for Health Metrics and Evaluation) kept running these models that predicted that the pandemic would end in 3 months, and then 3 months later would update their models (to predict that the pandemic would end in 3 months, again). 
  
The type of modeling used by the IHME, spline fitting, is in fact “a bunch of lines fitted to data points”, and it’s one of the trickiest classes of models to get right. Spline fitting has become more popular in recent years, and it underpins a number of machine learning (AI) approaches as well. 
  
One of the appealing aspects of spline fitting is that it’s “model agnostic”, which means that the models it uses are built from polynomials, a very broad and general class of functions. But this “agnosticism” comes at a cost – splines are easy to overfit. This weakness makes them often a poor choice when something is known about the underlying system, for example, when you are working with pharmacokinetic (PK) or pharmacodynamic (PD) data. 
  
In this white paper, we take a closer look  at spline fitting (using a set of simple polynomial functions) and explore what can go wrong with a “model agnostic” approach. While the white paper, on the surface, is about comparing two different kinds of modeling against each other, there are some deeper inferences to be drawn about the meaning of life. 
  
Just kidding. But there are deeper inferences that can be drawn about how we extract knowledge from a mathematical model, and what it really means (and doesn’t mean) when we say “the model fits the data”. 
  
So, there you have it- spline fits and the meaning of meaning. Grab yourself a snack and dig right in! 
 
#ModelFitting #SplineFitting #AI #DrugDevelopment #PKPD 

OSF Preprints. February 8.

predicted

The U.S. is in a sixth COVID wave—but it doesn’t look like it on a new CDC map

SDSU professor improves COVID-19 prevention spray protocol - Brookings Register

Total U.S. cases of COVID could more than double late this year, White House warns

work

Will COVID ever end? A forgotten pandemic from the late 1800s might offer some clues

An interview with Arijit Chakravarty on ending of the COVID-19 Public Health Emergency: Part 1 - WSWS

medRxiv 2022.01.26.22269854

05/2021

PLoS One. 2021 Mar 25;16(3):e0248509.

Not taking precautions? Be prepared to get COVID once a year, modeling shows

World Health Organization falsely declares end of COVID-19 public health emergency - WSWS

Population pharmacokinetics (popPK) modeling is a cornerstone of modern drug development, providing critical insights that drive decision-making from preclinical stages through to clinical trials. However, the process of building a reliable popPK model is fraught with potential pitfalls that can undermine its utility. In our latest white paper, we dig into these challenges in depth, offering practical guidance on how to navigate them effectively. 
 
At its heart, popPK modeling involves a series of critical decisions, from selecting the model structure to determining which parameters should account for interindividual variability. The choices made at each step can significantly impact the model's predictive power and its ability to support key decisions, such as dose selection and patient population extrapolation. 
 
One of the most significant risks in popPK modeling is overparameterization, where the model becomes too complex, incorporating more parameters than the data can reliably support. Overfitting is a related concern, where the model begins to capture noise rather than true signal, leading to poor predictive performance. We discuss how to mitigate these risks in this white paper, during both the design and evaluation stages.  
 
Another pitfall is poor covariate model selection. The inclusion of too many or poorly chosen covariates can introduce collinearity, skewing parameter estimates and reducing the model's reliability. In the white paper, we offer practical advice on selecting covariates with clinical or mechanistic significance and building the covariate model in a stepwise manner. The white paper also addresses the challenge of choosing appropriate distributions for random effects. While lognormal distributions are often used, they may not always be suitable. Visual tools like QQ plots and visual predictive checks (VPCs) are invaluable for assessing the appropriateness of distribution choices and ensuring the model accurately captures variability in PK data. 
 
Ultimately, a well-constructed popPK model is a powerful tool that enhances rational decision-making throughout the drug development process. By avoiding common pitfalls and focusing on clinical relevance and parsimony, developers can build models that support robust, data-driven decisions from early-stage development through to regulatory approval. 
 
If you’re involved in PK modeling – either as a producer or a consumer – and want to ensure your models are both robust and clinically relevant, click below to read the full white paper and gain insights into optimizing your modeling strategies.

#PopulationPK #DrugDevelopment #Pharmacometrics

Fractal Therapeutics

Business:

About: