Artificial Intelligence for RNAi Gene Target Discovery Still Early

It’s not every day that you can see a person like chess playing legend Gary Kasparov at a biotech company’s Investors Day.  And if you did, you might wonder whether this was  some kind of stunt.

I was therefore pleasantly surprised to find that Kasparov was indeed a very fitting appearance investor event by Silence Therapeutics this week (for presentation here).  In the 80s, Kasparov apparently gained an edge over the competition by analyzing opponents’ playing habits using ‘big data’ at a time that personal computers were only about to take off.  As a former professional chess player himself, the CEO of Silence (Ali Mortazavi) therefore believes that there similarly must be ways to more efficiently take advantage of the realms of genomic data being generated to gain an edge over the pharmaceutical competition in finding attractive targets and indications for RNAi gene silencing.

Herd mentality

One of the main reasons I used to look down at Big Pharma  with disdain was their herd mentality especially with regard to the targets and indications they are pursuing.  You can bet that if there is general hype around a certain target, or that once a target has been validated for a commercially attractive indication in a clinical proof-of-concept study, almost all Big Pharma companies will be pursuing corresponding programs.

Unfortunately, such herd mentality is not restricted to Big Pharma, but can also be observed in the biotech space (see PD/PD-L1 for cancer etc etc), even in the subsector dearest to my heart: oligonucleotide therapeutics.  Be it TTR, AAT, DMD, or HBV: once there is a compelling therapeutic rationale for pursuing a gene target, multiple companies will be on its case.

It is for this reason, particularly the fear that an Alnylam or Ionis will scorch the earth around these targets, that many companies now delay disclosing the nature of their most promising preclinical programs.  This is remarkable since small biotechs like Dicerna typically rely on disclosure of their of these programs to garner the necessary investor interest.

Target scarcity?

The fact that the lead programs of RNAi therapeutics companies frequently overlap also beckons the question of whether there is a biological scarcity of targets available.  This would be in contrast to Alnylam which have limited themselves to GalNAc-RNAi trigger conjugates for gene knockdown in the liver, claiming- in the flowery languish of its CEO- to be ‘drinking from a firehose of opportunities’ in just the hepatocyte.

To Alnylam’s credit, it has not only duplicated some of the programs initiated by antisense competitor Ionis Pharmaceuticals and that of its smaller RNAi competitors, they have done some of the heavy lifting themselves.  It was e.g. Alnylam that realized the true value of going after transthyretin for TTR amyloidosis and it was Alnylam that selected a target as unorthodox, but promising as antithrombin for the treatment of a protein deficiency: hemophilia (à ALN-AT3/Fitusiran). 

Other companies like Dicerna and Arrowhead Research can also be complimented for unearthing gene target nuggets for orphan indications like primary hyperoxaluria and liver disease due to certain mutations underlying alpha-1-antitrypsin disease.

Mining for targets

I am not concerned about a dearth of suitable targets that could inhibit the continued growth of RNAi Therapeutics.  This is because there are thousands of rare and severe diseases for which there should be straight-forward genetic solutions and because genetic/genomic information continues to explode.  Still, there is certainly tremendous value if you are the first to gain high conviction around a new target and it was the CEO of Silence who bemoaned what seemed to him like an archaic, manual process of sifting through the genetic ideas one by one.

The panel discussion largely cautioned that human curation will remain dominant for the foreseeable future.  I am slightly more optimistic (or pessimistic, depending on your attitude towards AI) and give AI 5 years or so until it will become a more and more compelling means for driving gene target discovery.  This delay relative to areas like the internet is largely explained by the fact that whereas genomic information has exploded, databases that link them to careful, systematic medical phenotyping are still in their infancy. 

In the meantime, some low-hanging fodder for oligonucleotide drug development may come as a by-product of genetically diagnosing rare diseases when often the last hope for getting a grip on a condition is to genome sequence the patient.

There are beneficiaries already from the recognition of the importance of pairing phenotype with genotype databases such as 23andMe which is said to be getting quite a bit interest from pharmaceutical companies and investors. This comes after the pioneer of sifting through genomic/phenomic data for target discovery, deCODE of Iceland, was acquired by Amgen in 2012.  deCODE rose to fame by linking population-wide genetic information to national health-record databases.  I would think that we will see quite a few more similar endeavors linking existing (à nations with a high level of social bureaucracy such as the Nordic countries) or IT-driven newly generated (à Google, wearables) health records to genome and transcriptome databases. 

And as these databases grow and become more intelligent, it will be those that first understand to harness them that will gain a Kasparov-style edge over the competition.  But don’t be afraid: AI should not make human gene target discoverers/evaluators redundant any time soon and their cost relative to the cost of developing drugs is too low.  If anything, they will be needed to make the ultimate decision of whether to pursue a proposed target as each disease is unique and it will take a long time until enough billion-$ clinical experiments will have been run to provide sufficient feedback for the AI to improve.

 Silence Therapeutics: Ready for Take-off

I usually do not attend biotech companies’ Investors Days, but since London is not too far away from home, I took advantage of the proximity to get a feel for how the company has matured. 

Silence Therapeutics has long been a player in the area of RNAi trigger design and IP (see also their claim on Alnylam products), but has struggled to come up with compelling therapeutic programs.  It was as if they never really tried.  This can probably be attributed to them having lacked the personnel with the experience of taking an idea through the clinic and onto the market.

It now seems as if they not only got the technology (à focus on GalNAc-RNAi) to the point of clinical maturity, but also have assembled the management with the skill and will to succeed in developing game-changing RNAi drugs.  First up will be SLN124 targeting TMPRSS6 for iron overload disorders with a projected CTA/IND filing in Q4 2018.