Executive Summary

1. Problem: Off-target interactions of drugs in clinical trials are not studied enough, resulting in a poor understanding of their mechanism of action. This
2. Opportunity: More effective target validation (in tandem with earlier proof-of-concept studies) could reduce attrition in phase II clinical trials by 24%, lowering cost of developing new drugs for novel MoAs by 30%. Better understanding of on- and off-target effects could be used to discover biomarkers, which could in turn be used for patient screening, patient stratification, and repurposing of drugs. The use of biomarkers would also triple LOA.
3. Solution: We're making novel leaps in target assessment and validation for oncology drugs to mitigate any unforeseen off-target effects early on in the drug development pipeline. By using machine learning to parse several papers and previous clinical trials, we can not only identify connections between different mechanisms of actions, cell lines, and off-target effects, but also propose key changes and recommendations for designing a trial that address other key problems such as patient screening by providing novel biomarkers for positive patient collection, repurposing a drug based on its off-target for novel usage, or simply learning more abut the drug prior to the trials.

Effective target validation along with previous proof of concept studies can reduce attrition in phase II clinical trials by 24%. The result of this would lower the cost of developing new drugs for novel MoAs by 30%. An in-depth understanding of on and off target effects can be used to discover biomarkers which could be used for patient screening, patient stratification, and the repurposing of drugs, additionally tripling LOA.

By making novel leaps

Outlining Status Quo

• RNAi studies are commonly used to validate targets. These studies use siRNA and shRNA to selectively knockout genes that have been identified as potential targets. The subsequent death of cancer cells treated with RNAi would indicate that the target identified is indeed a potential target.
• However, studies have shown that this method is imperfect, since the process of using RNAi can damage cancer cells before the target gene is knocked out.
• In most cases the time-pressure prevents the drug and its mechanism from having a strong genetic foundation which can directly explain the inner-workings, most of this is close-boxed thanks to RNAi studies and the lack of investigation and observation of its effects in the tumor or tissue sample itself (ie. studies only look at the cell's survival mechanism, not any other kinds of extra effects such as suppression of anti-cancer immune response)

"In principle, there is no regulatory requirement to know the molecular target of a drug or clinical candidate since all that matters in the end is that a drug is safe and efficacious. In fact, there are a number of approved drugs for which the mechanism-of-action is unknown. However, the drug development process is obviously greatly facilitated if the target is known since this enables rational design of new molecules with improved potency and safety profiles." Dr. Kilian V. M. Huber, Structural Genomics Consortium & Target Discovery Institute, University of Oxford.

Case Study on why the problem is big

1. 29 clinical trials based on inhibitors whose MoA turned out to be offset targets (HDAC6, MAPK14/p38α, PAK4, PBK, and PIM1
   1. Based on RNAi studies, these five genes were considered essential for cancer survival, and thus viable targets. However, a new study using CRISPR for gene knockout indicates that this is not the case. Further testing confirmed that this disagreement between RNAi results was not because of accidental mutations caused by CRISPR.
   2. Further testing also confirmed that knocking out these genes does not make cancer more vulnerable to chemotherapy.
   3. However, the drugs that were initially designed to fight cancer by inhibiting these genes are actually effective against cancer, proving that their mechanism of action is entirely different from what was previously thought.
2. Case Study of OTS964 (Same paper)
   1. OTS964 is a drug that was designed to fight cancer by inhibiting a gene called PBK. However, as mentioned earlier, PBK is not actually a cancer dependency. The study shows that the drug works by targeting a kinase called CDK11. This is especially significant since no drugs have previously been reported to target CDK11.
3. Case Study of OTS167, a MELK inhibitor that kills cancer through off-target effects
   1. Maternal Embryonic Leucine Zipper Kinase (MELK) is a protein that was previously considered essential in multiple cancer types. However, this study found that knocking out MELK with CRISPR had no effect on the fitness of cancer cells. Despite this, OTS167 still continued to kill cancer cells with MELK knocked out, which was surprising since inhibiting MELK was OTS167's mechanism of action. This suggests that the drug functions via off-target effects.
4. Sunitinib cardiotoxicity ?

Connect Case Study Examples to Greater Industry

1. These case studies highlight the importance of studying off-target effects and the MoAs of drugs in clinical trials.
2. This lack of understanding of drug candidate MoAs is likely significant contributor to the high failure rates of clinical trials, since misidentifying a drug’s mechanism of action (MOA) could hamper efforts to uncover a biomarker capable of predicting therapeutic responses.

Clearly Highlight Potential Opportunity

How does opportunity relate back to problem statement

Why hasn't opportunity been seized yet?

Case Study of Astra Zeneca's and how they made money through the process of off-target effects and repurposing the mechanism of action

How can we leverage BenchSci

1. Customer Base → do they require this service or will we need to get new customers, and if so how?
   1. The importance of target validation and studying off-target effects is applicable to all pharma companies in drug discovery.
2. What IP does BenchSci have or resources to support this idea?
3. Why is it a logical step for BenchSci to pursue this and what gaps need to be filled?
   1. reason 1 is that BenchSci is expanding their portfolio of clinical agents to cell lines, reagents, CRISPR cell lines, animal models, PCR, and quite literally everything that is fundamental to drug discovery → target validation is a critical next step which not only aligns with some of the more fundamental ML tech you are using as of now, but is built on top of the findings that BenchSci showcases to its customers (i.e. evaluating off-targets based on cell lines and their respective differential expression promoters is the first step in identifying potential up/down regulation of on/off-target effects, all the data and proprietary tech is already on the platform)
4. Can this easily integrate into the platform, if so how?
5. Does this help support BenchSci's goal of getting novel medicine to patients 50% faster by 2025?
6. How does BenchSci win? What is their potential for growth?

Clear end goal with timeframe → if we implement this idea, some ideas of a goal could be capture __ % of oncology market of failed drugs by 2025 or save our customers up to __ million dollars in clinical and preclinical stages

Our solution directly expands upon the goal of BenchSci by focusing efforts in preclinical stages into useful findings for Clinical → draw this connection of feedback loop and the idea of making the improvements in preclinical much more prominent and impactful in clinical here (this is why our solution stands out, because its the most direct way to effect not just clinical POS but also preclinical)

Key difference than traditional approach used by BenchSci → while scientists have to traverse said knowledge graphs to find relationships, here we make clearer recommendations that are more useful for drug companies conducting clinical trials, further integrating both preclinical and clinical and using BenchSci to inform decisions across the entire pipeline

1. earlier decisions in preclinical that translate to clinical success improves probability of success by greater amounts than later on or in stage-stage jumps (each phase gets more expensive, changes in between make this heightened regardless if the goal is to improve some sort of logistical problem)

Case Study: Potential, actionable areas of implementation