1 of 17

�Recursion Pharmaceuticals:

Powerful Platform and Partnerships

PBG Biotech Investment Research Program

April 2023

James Glazar

2 of 17

Recommendation – Buy RXRX

2

I recommend buying RXRX
Valuation: $10 / share, currently trading at $5 / share
Catalysts: REC-3964 and REC-994 data within next 18 months.
Risks

Current drugs could be shown ineffective
Further partnerships may not materialize

However, I am assured by:
New directions in drug searching
Large investments from big pharma companies

3 of 17

Recursion screens drugs to accelerate development

3

Pipeline – Precision oncology, rare diseases

Partnership – Leverage expertise to speed discovery

Data – License access to subsets of data

Dean Li, MD, PhD

Co-founder

Chris Gibson, PhD

Co-founder, CEO

100%

80

60

51

100%

25

64

60

12

8

Screen

Hit

Validate lead

Advance

Develop

Industry avg.

Recursion

$ to IND

8M

22M

months to lead validation

30

10

subtle clues of rescue picked up by ML!

Gibson et al., Circulation (2015)

searching wider

failing faster

Gibson PhD work in Bray et al., Nature Protocols 2016 – dissertation in high-throughput cell treatment and imaging

Li works mainly on endothelial issues in vasculature

Li is current president of Merck Research Labs.

Also founded Hydra Biosciences in 2001 (acquired by Eli Lilly in 2018 for $22.6M) and Navigen Pharma in 2007 (has D-peptide inhibitor called CPT31 that jams HIV’s infection machinery. The molecule is now in phase 1 clinical trials)

Also co-founded with Blake Borgeson (another PhD student) who now works on board at Machine Intelligence Research Inst.

Dissertation was mainly on cerebral cavernous malformation (leaky endothelium in brain vasculature)

Tried hypothesis-driven testing with statins, but failed

Much of the work involved imaging as shown above – healthy vs diseased (CCM2 gene knockdown)

New idea – data-driven testing. Screen 2100 drugs via imaging → screen 38 compounds via electrical impedance test → screen 7 compounds via mouse dermal test → try 2 compounds (Tempol and Vitamin D – both work!)

Only 1 expert-picked drug works in impedance test, whereas 8 ML-picked drugs worked – ML is better at picking the winners in high-dimensional space

Scale bars are 50 microns

green = actin, blue = DNA, red = VE-Cadherin (vascular endothelial cadherin)

I think the ML picked up on colocalization / linearity of actin and VE-cadherin

500 employees – 40% bio, 35% data, 25% operations

test drugs, then take images (cheap and data-rich) for pennies per sample. Measure size, shape, texture, and spatial distances

See if drugs make diseased cells appear more healthy

Break genes then see how images change; Mine data to find genes with similar effects

$ to IND is main value proposition – make drug approval overall cheaper by failing fast. Potential to severely lower drug costs

Industrial program generation; hypothesis validation via transcriptomics and genomics (automated)

Focus on small molecule for now, can try biologics later

21Pb of images, proteomics, genomics, metabolomics

Open source teasers (some data blanked out)

4 of 17

Partnerships drive value and provide stable backbone

4

Recursion offers:

Wider screening net with higher hit quality
Distinct approach compared to competitors

mature pipeline
advanced computation on-site
wet+dry lab integration

Maps of relevant biological data

11/2013

Founded

6/2021

Invest in Enveda

7/2020

Acquire Vium

4/2021

IPO - $436M

9/2020

Bayer Partnership

12/2021

Roche Partnership

Fibrotic disease

$30M upfront + $50M equity

$100M per program for up to 12 programs

8 total programs initiated

Neuroscience and oncology

$150M upfront

$300M per program for up to 40 programs

Partners offer:

Proprietary datasets
Subject matter expertise
Production and distribution capabilities

successful listings of biotech like 10x Genomics and Relay Therapeutics caused VC investment in AI-enabled computational biology to nearly triple in 2020 to over $3.2 billion

2014 – $170k seed

2020 Vium acquisition for undisclosed amount, likely around $100M

Vium founded by Joe Betts-Lacroix – inventor at MIT and Caltech, started OQO to make small PCs (acquired by Google), then Vium to make image/data collecting mouse cages (acquired by RXRX in 2020), now launched Retro Biosciences with $180M from Sam Altman in 2022.

2021 Enveda investment <$50M for sure, likely around $5M

Enveda founded by Viswa Colluru – Immuno-oncology PhD at U. Wisconsin-Madison, then joined Recursion in 2016 as scientist, left in 4/2019

knowledge graph ML for plant metabolomics; Google for drug discovery searches

Bayer – Fibrosis, $30M upfront, $50M equity, $1.2B for 12 programs, they get ~5% cut royalties

Started brute force on small molecules to see fibrosis improvements, now inferential (ML-guided) search

5 year term; 500k phenoprints done

Roche/Genentech – Neuroscience + Oncology, $150M upfront, $300M per program for 40 programs, ~8% royalties

Up to $250M for 16 accepted Phenomaps

CEO Aviv Regev – computational biologist

Many big pharma companies are looking for late-stage (I’d imagine esp. due to rising interest rates)

Market cap of company is ~$1B

4 search modalities – brute force novel chemicals (REC-3694, Bayer fibrosis programs), brute force known chemicals (REC-994, REC-2282, REC-4881), mapping/navigating novel chemicals (REC-65029, REC-14221, Genentech programs), mapping/navigating known chemicals (REC-2029)

big pharma likes:

innovative first-in-class or best-in-class assets for unmet medical need

demonstration of target engagement in vivo.

novel chemical entity composition of matter patent protection

back-up molecules

worldwide rights

no indication splitting – licensing one indication to one party and another to another party.

5 of 17

Catalysts – 1 moderate, 1 big

5

REC-994 – Cerebral Cavernous Malformation

Phase 2 results

360k symptomatic patients in US+EU

Est. 1M total patients

20% have genetic link, which is more severe

Leads to hemorrhage, seizure, stroke

Orphan drug – surgery only for accessible lesions

Est. peak revenue / year – $1.5B

2H 2023

REC-3964

REC-3964 – C. difficile Colitis

Phase 1 results

Common, with 750k cases / year

120k persistent cases

40k elderly die / year

Leads to diarrhea, then colitis and sepsis

3 competitor drugs, 2 FDA-approved

Est. peak revenue / year – $300M

2024

REC-2282

2H 2024

REC-994

(Phase 2, rare) CCM – REC-994 (Tempol); life expectancy is 55 years; Phase 1 (32 patients at 50-800mg, passed) Phase 2 (60 patients at 200/400mg)

(Phase 2, rare) Type 2 Neurofibromatosis – REC-2282 – POC interim readouts for safety analysis in 2024 (not a big deal); tumors on nerves; mutations in NF2 tumor suppressor gene; hearing loss, balance issues; in-licensed from Arno Therapeutics after Recursion’s statistical validation; drug shows tumor slowdown; Phase 2/3 combined study, with Phase 2 interim safety readout in 2024; ~$220M est. peak revenue / year in US

(Phase 1, rare) Clostridioides difficile Colitis – REC-3964 – 1st Recursion-only small molecule; Common after post-surgery antibiotics; 56 SAD, 50 MAD patients;

(Phase 2, onco) Familial Adenomatous Polyposis – REC-4881, has fast-track designation

(Phase 1, onco) AXIN1/APC mutant cancers – REC-4881, additional INDs and clinical starts, including Ph1b/2 trial initiation for AXIN1/APC programs

(Pre-clinical, onco) Cancer immunotherapy target delta

(Late disc, onco) HR-proficient ovarian cancer RBM39 – REC-65029

(Late disc, onco) Cancer immunotherapy target alpha – REC-648918

(Late disc, onco) MYC-driven oncology – REC-14221

6 of 17

Drugs and partnerships propel revenues starting 2027

6

Analyst	Target ($/share)	Notes
SVB Securities	9	Dropped from $33 after Ph2 REC-4881 delay and shelving REC-3599 for 2 years
JPMorgan	17	Predicts $500M from REC-2282
Needham	17	Predicts high partnership revenue in mid-2030s
Berenberg	33	Predicts peak sales of $800M for each drug

77 buy, 22 hold, 0 sell

🡪 apply 10% discount rate and 3% terminal growth rate 🡪 $10 / share

	2023	2024	2025	2026	2027	2028	2029	2030	2031	2032	2033
partners	50	100	150	150	300	350	400	400	450	500	500
REC-994	-	-	-	-	100	300	400	500	550	500	500
REC-4881	-	-	-	-	100	200	300	350	350	300	300
total risk-adj revenue	50	50	100	100	500	850	1100	1250	1350	1300	1300
costs	300	350	350	400	450	485	510	525	535	530	530
earnings	-250	-300	-300	-300	50	365	590	725	815	770	770

My simple model doesn’t account for:

smaller drugs in pipeline
newly developed drugs
accelerated partnership development

Risk adjustment is ~30% prob. of success

Taxes of ~20% kick in with drugs (I use 10%)

Research expenses cap out around $400M – lab already producing Pb of data

REC-3599 for ultra-rare indication infantile GM2 gangliosidosis pushed by 2 years to do sheep model studies before infants

Scrapped in October 2022 to put resources towards REC-4881 (FAP)

SVB slashed valuation from $33 to $10 in March 2022 – halved prob. of success for all programs and pushed out by 1 year

Discount rates (%) – SVB 12, JPMorgan 13, Needham 15, Berenberg 13

Terminal growth rates (%) – SVB 0, JPMorgan 3, Needham 3, Berenberg 3

$550M cash + cash equivalents, $280M per year operating cost, $45M revenues

Aside – RXRX has price/book of ~2 so the worst that happens is it halves (assuming cash remains constant and doesn’t grow)

High quick ratio indicates ability to pay short-term debt

RXRX – price/book = 1.95 – quick ratio = 5.5

RLAY – price/book = 1.45 – quick ratio = 15.8

ABCL – price/book = 1.57 – quick ratio = 8.6

TWST – price/book = 0.95 – quick ratio = 5.8

high discounting of future cash flows

don’t rely on acquisition

check financial ratios against peers

higher partnership revenue than competitors

7 of 17

Risks – drugs failing, lack of new partnerships

7

Drugs could fail, indicating flaws in the platform

However, most drugs in pipeline were found via brute force (old method)

2. Current partnerships could expire with no new partnerships on the horizon

However, Bayer expanded 2 mapping programs one year after starting collaboration

Brute force new chemicals REC-3684 Bayer fibrosis programs
Brute force known chemicals REC-994 REC-2282

new search modality

old search modality

Map and navigate new chemicals Genentech programs
Map and navigate known chemicals REC-2029

8 of 17

Summary & Recommendations

8

Recursion has a vast data-driven platform for expedited drug screening

Current drugs include REC-994 for CCM and REC-4881 for FAP in Phase 2

Current partnerships include Bayer and Genentech – potential for future growth

I value RXRX at $10 / share – Buy

9 of 17

Appendix

9

Pharma success rate references

1 Alacrita Consulting. Pharmaceutical Probability of Success. (2018)�2 Deloitte. Ten years on: Measuring the return from pharmaceutical innovation (2020)�3 DiMasi et al. Innovation in the pharmaceutical industry: New estimates of R&D costs. Journal of Health Economics. 47:20-33 (2016)�4 Paul, et al. How to improve R&D productivity: the pharmaceutical industry’s grand challenge. Nature Reviews Drug Discovery. 9: 203-214 (2010) 5 Martin et al. Clinical trial cycle times continue to increase despite industry efforts. Nature Reviews Drug Discovery. 16:157 (2017)

10 of 17

REC-994 Pathway

10

11 of 17

REC-4881 Pathway

11

12 of 17

REC-2282 Pathway

12

13 of 17

REC-3964 Pathway

13

14 of 17

XBI and RXRX correlation

Returns (%)

XBI

RXRX

15 of 17

Partnerships

16 of 17

Partnerships

17 of 17

Valuation – Drugs and partnerships propel revenues starting 2027

17

Analyst	Target ($/share)	Notes
SVB Securities	9	Dropped from $33 after Ph2 REC-4881 delay and shelving REC-3599 for 2 years
JPMorgan	17
Needham	17	High partnership revenue in 2030s
Berenberg	33

77 buy, 22 hold, 0 sell

Ticker	Price/Book	Quick Ratio
RXRX	1.95	5.5
RLAY	1.45	15.8
ABCL	1.57	8.6
TWST	0.95	5.8

Worst case price = $2.50, Can repay 5.5x debt

	2023	2024	2025	2026	2027	2028	2029	2030	2031	2032	2033
partners	50M
REC-994
REC-4881
risk-adj revenue

Risk adjustment is ~30% prob. of success

2023 – partnerships = 50M

2024 – partnerships

2025 – partnerships + 50M

2026 – partnerships

2027 – partnerships, REC-994 = 50M, REC-4881 = 50M

2028 – partnerships + 50M, REC-994 + 50M, REC-4881 + 50M

2029 – partnerships + 50M, REC-994 + 100M, REC-4881 + 50M

2030 – partnerships + 50M, REC-994 + 100M, REC-4881 + 50M

2031 – partnerships + 50M, REC-994 + 100M, REC-4881 + 50M

2032 – partnerships + 50M, REC-994 + 100M, REC-4881 + 50M

2033 – partnerships + 50M, REC-994 + 50M, REC-4881