Digital Evolution at Wafer Scale

June 11, 2026 @ ERA Town Hall

​

Matthew Andres Moreno

Complex Systems/Ecology and Evolutionary Biology

University of Michigan

these slides: hopth.ru/gq

Luis Zaman

Emily Dolson

Matthew Andres

Moreno

Joey Wagner

Connor Yang (UROP)

Vivaan Singhvi (UROP)

Evolution Models in vivo and in silico

LTEE

(Good et al., 2017)

📧 morenoma@umich.edu

E.

coli

experimental evolution:

n>1, experimental manipulations

Evolution Models in vivo and in silico

LTEE

(Good et al., 2017)

Avida

(Ofria and Wilke, 2009)

​

📧 morenoma@umich.edu

E.

coli

experimental evolution:

n>1, experimental manipulations

​

simulation and modeling:

synthesize theory, test sufficiency

Evolution Models in vivo and in silico

LTEE

(Good et al., 2017)

Avida

(Ofria and Wilke, 2009)

​

📧 morenoma@umich.edu

E.

coli

1

processor

~billion replications/day

1

processor

1

processor

e.g.,

digital

multicell

experiment

morph

phenotype

morph

phenotype

stint

0

1

2

14

15

39

45

stint

59

74

100

@MorenoMatthewA

stint 0

morph a

@MorenoMatthewA

stint 14

morph d

@MorenoMatthewA

stint 15

morph e

@MorenoMatthewA

stint 45

morph g

1

processor

​

​

​

​

Cgoodwin, CC BY-SA 4.0 <https://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons

Nejones1987, CC BY-SA 4.0 <https://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons

a bigger ox

a team of oxen

✓

Grace Hopper

KLAT2 (HANK DIETZ)

KLAT2 (HANK DIETZ)

Fugaku

KLAT2 (HANK DIETZ)

The Library of Congress via Wikimedia Commons

“next-generation” high-performance computing (especially AI/ML)

KLAT2 (HANK DIETZ)

The Library of Congress via Wikimedia Commons

NVIDIA A100 GPU

6,912 CUDA cores

KLAT2 (HANK DIETZ)

The Library of Congress via Wikimedia Commons

Graphcore IPU

1,200 cores per chip

clustered up to 1,024

chips

​

NVIDIA A100 GPU

6,912 CUDA cores

SambaNova

RDU

KLAT2 (HANK DIETZ)

Cerebras

Wafer-Scale

Engine

The Library of Congress via Wikimedia Commons

850,000

​

​

cores

NVIDIA A100 GPU

6,912 CUDA cores

Graphcore IPU

1,200 cores per chip

clustered up to 1,024

chips

​

SambaNova

RDU

Cerebras Wafer-Scale Engine

850,000

cores

📧 morenoma@umich.edu

Cerebras Wafer-Scale Engine

850,000

cores

📧 morenoma@umich.edu

…

…

…

…

…

…

…

…

Cerebras Wafer-Scale Engine

850,000

cores

📧 morenoma@umich.edu

​

​

…

…

…

…

…

…

…

…

Cerebras Wafer-Scale Engine

850,000

cores

📧 morenoma@umich.edu

​

​

…

…

…

…

…

…

…

…

Cerebras Wafer-Scale Engine

850,000

cores

📧 morenoma@umich.edu

​

​

data

output

…

…

…

…

…

…

…

…

Cerebras Wafer-Scale Engine

850,000

cores

📧 morenoma@umich.edu

​

​

only ~48kb

mem per core

GraphCore IPU — another AI/ML accelerator

1,200 cores per chip

clustered up to 1,024 chips

Mapping Evolution Simulations onto WSE

Treatment Conditions. Allowed mutational outcomes under compared outcomes. Note that adaptive regime introduces the possibility for selective sweeps.

a) purifying

regime

b) adaptive

regime

Example reconstructions. Phylogenetic reconstructions of 1 million generation on-hardware simulations. For legibility, phylogeny visualizations were further subsampled to 1k end-state agents. Left phylogenies are log-scaled with ultrametric correction to better show topology and right phylogenies are linear-scaled.

On-device Evolution Trial

Phylometric outcomes. Statistics calculated from reconstruction of 18 million population size/1 million generation on-hardware simulations. Phylometrics were calculated from reconstructions with 10k sampled end-state agents.

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

​

“

”

Mapping Evolution Simulations onto WSE

Treatment Conditions. Allowed mutational outcomes under compared outcomes. Note that adaptive regime introduces the possibility for selective sweeps.

a) purifying

regime

b) adaptive

regime

Example reconstructions. Phylogenetic reconstructions of 1 million generation on-hardware simulations. For legibility, phylogeny visualizations were further subsampled to 1k end-state agents. Left phylogenies are log-scaled with ultrametric correction to better show topology and right phylogenies are linear-scaled.

On-device Evolution Trial

Phylometric outcomes. Statistics calculated from reconstruction of 18 million population size/1 million generation on-hardware simulations. Phylometrics were calculated from reconstructions with 10k sampled end-state agents.

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

“

”

🐥🦜🦜🦜🦤🦤🐥

Mapping Evolution Simulations onto WSE

Treatment Conditions. Allowed mutational outcomes under compared outcomes. Note that adaptive regime introduces the possibility for selective sweeps.

a) purifying

regime

b) adaptive

regime

Example reconstructions. Phylogenetic reconstructions of 1 million generation on-hardware simulations. For legibility, phylogeny visualizations were further subsampled to 1k end-state agents. Left phylogenies are log-scaled with ultrametric correction to better show topology and right phylogenies are linear-scaled.

On-device Evolution Trial

Phylometric outcomes. Statistics calculated from reconstruction of 18 million population size/1 million generation on-hardware simulations. Phylometrics were calculated from reconstructions with 10k sampled end-state agents.

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

​

“

”

🧬🧬🧬🧬

​

[u32, u32…]

🐥🦜🦜🦜🦤🦤🐥

Mapping Evolution Simulations onto WSE

Treatment Conditions. Allowed mutational outcomes under compared outcomes. Note that adaptive regime introduces the possibility for selective sweeps.

a) purifying

regime

b) adaptive

regime

Example reconstructions. Phylogenetic reconstructions of 1 million generation on-hardware simulations. For legibility, phylogeny visualizations were further subsampled to 1k end-state agents. Left phylogenies are log-scaled with ultrametric correction to better show topology and right phylogenies are linear-scaled.

On-device Evolution Trial

Phylometric outcomes. Statistics calculated from reconstruction of 18 million population size/1 million generation on-hardware simulations. Phylometrics were calculated from reconstructions with 10k sampled end-state agents.

“

”

✉️🧬

🧬🧬🧬🧬

​

[u32, u32…]

🐥🦜🦜🦜🦤🦤🐥

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

​

Performance

WSE ~8x10¹¹ (800 billion) agent-generations per second

​

295× vs GPU

111,091× vs CPU

Goal: develop and apply methods to harness next-generation high-performance computing hardware to enable larger agent-based evolution simulations

WSE

Goal: develop and apply methods to harness next-generation high-performance computing hardware to enable larger agent-based evolution simulations

WSE

Goal: develop and apply methods to harness next-generation high-performance computing hardware to enable larger agent-based evolution simulations

WSE

This Talk: methodology for collecting evolutionary history from wafer-scale evolution simulations

…

…

…

…

…

…

…

…

Cerebras Wafer-Scale Engine

48kb mem

😶‍🌫️

📧 morenoma@umich.edu

…

…

…

…

…

…

…

…

Cerebras Wafer-Scale Engine

48kb mem

😶‍🌫️

🌸

🌺

🌼

🌻

phylogeny

📧 morenoma@umich.edu

What is a phylogeny and what can it tell you?

🍞🧈

What is a phylogeny and what can it tell you?

Simulation Phylogeny Data:

1. history of evolutionary events
2. insight into evolutionary dynamics
3. “ground truth” data to test/develop bioinformatics tools

What is a phylogeny and what can it tell you?

Root

Leaves

(aka Tips)

What is a phylogeny and what can it tell you?

closely related

Lineages diverged

What is a phylogeny and what can it tell you?

closely related

“fossil tips”

What is a phylogeny and what can it tell you?

🦕

What is a phylogeny and what can it tell you?

=

=

Asexual Phylogeny

vertex:

“taxonomic unit”

🦠

Asexual Phylogeny

vertex:

“taxonomic unit”

🦠

🦠

Asexual Phylogeny

vertex:

“taxonomic unit”

🦠

🦠

🦠

🦠

Asexual Phylogeny

​

Organisms vs. Species

Sexual vs asexual phylogenies

Sexual vs asexual phylogenies

​

Sexual vs asexual phylogenies

​

Sexual vs asexual phylogenies

​

Sexual vs asexual phylogenies

Sexual vs asexual phylogenies

Sexual vs asexual phylogenies

​

Baum and Offner, 2008

Vs

long term coexistence

U.S. Navy, Public domain, via Wikimedia Commons

What is a phylogeny and what can it tell you?

Vs

Moreno et al., 2023

4 niche

ecology

recent selective sweep

Perfect Tracking:

Complete Observability

Perfect Tracking:

Complete Observability

(Nozoe et al, 2017)

🦠

🦠

🦠

🔬

🦠

Perfect Tracking:

Complete Observability

(Nozoe et al, 2017)

🔬

🦠

🦠

🦠

🦠

Perfect Tracking:

Complete Observability

🌸

🌺

🌼

🌻

🦠

🦠

🦠

👍

(Nozoe et al, 2017)

🔬

🦠

Perfect Tracking:

Complete Observability

🌸

🌺

🌼

🌻

👍

(Nozoe et al, 2017)

🔬

👎

🦠

🦠

🦠

🦠

Perfect Tracking:

Complete Observability

🌸

🌺

🌼

🌻

👍

👎

(Nozoe et al, 2017)

🔬

🦠

🦠

🦠

🦠

Costs of Perfect Tracking

​

👍

👎

🔬

• Memory, Storage, & Bandwidth
• e.g., 8PB vs 30GB
• Runtime Extinction Tracking Costs
• Hardware Crashout & Corruption
• perfect tracking is sensitive to data loss
• at exascale, >1 node fail/hour

WSE

Chip

host

PE row 410 between columns 0 and 248

~5% of 32-bit words from region corrupted

0.03% of overall extracted 5-word genomes

@MorenoMatthewA 🐘@mas.to

🌸

🌺

🌼

🪷

🌻

🔬

🦠

🦠

🦠

@MorenoMatthewA 🐘@mas.to

🌸

🌺

🌼

🌻

🌸

🌺

🌼

🪷

🌻

🔬

🦠

🦠

🦠

@MorenoMatthewA 🐘@mas.to

🌸

🌺

🌼

🌻

🌸

🌺

🌼

🪷

🧬

🧬

🧬

🧬

🌻

🔬

🦠

🦠

🦠

@MorenoMatthewA 🐘@mas.to

​

🪄🐇

​

🌸

🌺

🌼

🌻

🌺

🌼

🪷

🧬

🧬

🧬

🧬

🌻

🔬

🦠

🦠

🦠

@MorenoMatthewA 🐘@mas.to

🌸

🌺

🌼

🌻

🌸

🌺

🌼

🪷

🧬

🧬

🧬

*inferred estimate

🧬

🌻

🔬

🦠

🦠

🦠

@MorenoMatthewA 🐘@mas.to

🌸

🌺

🌼

🌻

🌸

🌺

🌼

🪷

🧬

🧬

🧬

🧬

*inferred estimate

How to design

.🧬 to facilitate reconstruction?

@MorenoMatthewA 🐘@mas.to

🌸

🌺

🌼

🌻

🌸

🌺

🌼

🪷

🧬

🧬

🧬

🧬

*inferred estimate

Hwang et al., 2019

Crispr-cas9 Barcoding

@MorenoMatthewA 🐘@mas.to

🌸

🌺

🌼

🌻

🌸

🌺

🌼

🪷

🧬

🧬

🧬

🧬

…

…

🧬

= “hstrat” annotation

@MorenoMatthewA 🐘@mas.to

🌸

🌺

🌼

🌻

🌸

🌺

🌼

🪷

🧬

🧬

🧬

🧬

= “hstrat” annotation

…

…

🧬

• small (~96 bits)
• fast (< 500ns)
• tunably accurate

evolve

under-the-hood “hstrat” algorithm

evolve

under-the-hood “hstrat” algorithm

(✂️ for time)

evolve

evolve

evolve

.B.

.C.

.A.

evolve

.B.

.C.

.A.

evolve

.C.

.B.

.A.

.B.

.C.

.A.

evolve

end-state

.C.

.B.

.A.

.C.

.B.

.A.

.B.

.C.

.A.

evolve

end-state

reconstruct

.C.

.B.

.A.

.C.

.B.

.A.

.B.

.C.

.A.

evolve

end-state

reconstruct

✂️

(q&a welcome)

t=8

t=8

t=8

t=16

t=8

t=16

t=8

t=16

t=8

t=16

t=8

t=16

“steady”

“tilted”

t=8

t=16

t=8

t=16

“steady”

“tilted”

This is the hard/interesting part!!!

Ackley, David H. "A robust programmable replicator for an indefinitely scalable machine." In Artificial Life Conference Proceedings 35, vol. 2023, no. 1, p. 81

https://doi.org/10.1162/isal_a_00701

​

Ackley, David H. "A robust programmable replicator for an indefinitely scalable machine." In Artificial Life Conference Proceedings 35, vol. 2023, no. 1, p. 81

https://doi.org/10.1162/isal_a_00701

​

data stream

data storage

Generalized Ring Buffer (Gunther, 2014)

data stream

data storage

It is generation 72…

… put at position 4

temporal coverage

Simple Ring Buffer

temporal coverage

Steady Retention

temporal coverage

Tilted Retention

Generalized Ring Buffer (Gunther, 2014)

data storage

72 generations have elapsed…

gen 68

gen 54

gen 45

gen 23

gen 61

Time

WSE Chip

Time

sample 0

WSE Chip

🧬samples

Time

WSE Chip

🧬samples

sample 0

Time

WSE Chip

🧬samples

sample 0

Time

sample 0

WSE Chip

🧬samples

Time

sample 0

sample 1

WSE Chip

🧬samples

Time

sample 0

sample 1

sample 2

sample 3

sample 4

sample 5

WSE Chip

⋮

​

Time

sample 0

sample 1

sample 2

sample 3

sample 4

sample 5

sample 6

Reconstructed Phylogeny

⋮

​

WSE Chip

⋮

​

Time

sample 0

sample 1

sample 2

sample 3

sample 4

sample 5

sample 6

Reconstructed Phylogeny

⋮

​

WSE Chip

some

data

loss ok

⋮

​

Time

sample 0

sample 1

sample 2

sample 3

sample 4

sample 5

sample 6

Reconstructed Phylogeny

⋮

​

WSE Chip

some

data

loss ok

1.13 quadrillion evaluations, 226 million pop size, 500,000 generations,

10 minutes runtime, 9,000 snapshots (15/sec), 0.0001% sampled, 1 billion tip phylogeny

hstrat Pipeline

1. raw sim “genome” strings

​

2. extract and decode

markers

3. build tree

4. phylogeny

courtesy meat-machinery.com

.C.

.B.

.A.

.C.

.B.

.A.

Connor

Yang

Vivaan Singhvi

Joey

Wagner

​

⇤ 1170 ⇥

⇤ 1170 ⇥

sample

genomes

WSE Chip

755×1170

PE array

⇤ 755 ⇥

sample

lineage

⇤ 1170 ⇥

sample

genomes

WSE Chip

755×1170

PE array

sample

lineage

⇤ 1170 ⇥

⇤ 1170 ⇥

sample

genomes

WSE Chip

755×1170

PE array

⇤ 755 ⇥

sample

lineage

directional

migration

bias

Neutral Conditions

Adaptive Conditions

Neutral Conditions

Adaptive Conditions

Conclusion

“perfect” observability

(Nozoe et al,

2017)

🦠

🦠

🦠

🦠

“sampling-based” observability

“perfect” observability

(Nozoe et al,

2017)

🦠

🌸

🌺

🌼

🌻

🧬

🧬

🧬

🧬

🦠

🦠

🦠

“sampling-based” observability

“perfect” observability

(Nozoe et al,

2017)

🦠

🌸

🌺

🌼

🌻

🧬

🧬

🧬

🧬

🦠

🦠

🦠

in vivo

“sampling-based” observability

“perfect” observability

(Nozoe et al,

2017)

🦠

🌸

🌺

🌼

🌻

🧬

🧬

🧬

🧬

🦠

🦠

🦠

in silico

in vivo

“sampling-based” observability

“perfect” observability

(Nozoe et al,

2017)

🦠

🌸

🌺

🌼

🌻

🧬

🧬

🧬

🧬

🦠

🦠

🦠

🔬

… ← in vivo

in silico

“sampling-based” observability

“perfect” observability

(Nozoe et al,

2017)

🦠

🌸

🌺

🌼

🌻

🧬

🧬

🧬

🧬

🦠

🦠

🦠

🔬

in silico → …

… ← in vivo

“sampling-based” observability

“perfect” observability

(Nozoe et al,

2017)

🦠

🌸

🌺

🌼

🌻

🧬

🧬

🧬

🧬

🦠

🦠

🦠

🔬

in silico → …

… ← in vivo

(Dolson and Ofria, 2021)

Parallel/distributed Scale-up: Interesting Trade-offs

• complete observability / sampling-based observability

(Dolson and Ofria, 2021)

Peter J. Park, CC BY 2.5 <https://creativecommons.org/licenses/by/2.5>, via Wikimedia Commons

kqedquest Via Flickr

Wafer-Scale Cluster

Peter J. Park, CC BY 2.5 <https://creativecommons.org/licenses/by/2.5>, via Wikimedia Commons

kqedquest Via Flickr

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

2,048

CS-3

​

Wafer-Scale Cluster

Peter J. Park, CC BY 2.5 <https://creativecommons.org/licenses/by/2.5>, via Wikimedia Commons

kqedquest Via Flickr

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

WSE Chip

⇤1170⇥

⇤755⇥

​

2,048

CS-3

​

1.74 B

processors

256

exaFLOPs

🧬

🧬

🧬

🧬

Wafer-Scale Cluster

Dr. Luis Zaman

@MorenoMatthewA 🐘@mas.to

Dr. Emily Dolson

📧 morenoma@umich.edu

Joey Wagner

Connor Yang (UROP)

Vivaan Singhvi (UROP)

Office of Advanced Scientific Computing Research (ASCR)

Award Number DE-SC0025634

ByteBoost Workshop ‘24

ByteBoost

Workshop ‘24

@MorenoMatthewA 🐘@mas.to

📧 morenoma@umich.edu

(Mei Yu, Julian, & Riaz)

(Extensive)

Technical Assistance

“Re: Re: Fwd: Troubleshooting…”

Questions?

We have a very bad tendency to base our plans for computers on the equipment we have in house and the things we’re doing now. And totally fail to review them in the light of the equipment that will be available and the things that we will be doing — I think the saddest phrase I ever hear in a computer installation is that horrible one "but we’ve always done it that way." That’s a forbidden phrase in my office.

Capt. Grace Hopper

Future Possibilities: Data, Hardware, Software, and People — August 26, 1982

these slides: hopth.ru/gq

📧 morenoma@umich.edu

References

• Ackley, David H. "A Robust Programmable Replicator for an Indefinitely Scalable Machine." ALIFE 2023: Ghost in the Machine: Proceedings of the 2023 Artificial Life Conference. MIT Press, 2023.
• Liquidware. Introducing the Illuminato X Machina. http://antipastohw.blogspot.com/2009/08/introducing-illuminato-x-machina.html
• ALIEN Artificial Life Environment. https://www.alien-project.org/index.html
• Dolson, Emily, and Charles Ofria. "Digital evolution for ecology research: a review." Frontiers in Ecology and Evolution 9 (2021): 750779.
• Good, Benjamin H., et al. "The dynamics of molecular evolution over 60,000 generations." Nature 551.7678 (2017): 45-50.
• Ofria, Charles, and Claus O. Wilke. "Avida: A software platform for research in computational evolutionary biology." Artificial life 10.2 (2004): 191-229.
• Buitrago P.A., Nystrom N.A. (2021) Neocortex and Bridges-2: A High Performance AI+HPC Ecosystem for Science, Discovery, and Societal Good. In: Nesmachnow S., Castro H., Tchernykh A. (eds) High Performance Computing. CARLA 2020. Communications in Computer and Information Science, vol 1327. Springer, Cham. https://doi.org/10.1007/978-3-030-68035-0_15

​

Problem Setup — [props?]

Problem Setup — [props?]

Proof-of-concept Experiment

Treatment Conditions. Allowed mutational outcomes under compared outcomes. Note that adaptive regime introduces the possibility for selective sweeps.

a) purifying

regime

b) adaptive

regime

Example reconstructions. Phylogenetic reconstructions of 1 million generation on-hardware simulations. For legibility, phylogeny visualizations were further subsampled to 1k end-state agents. Left phylogenies are log-scaled with ultrametric correction to better show topology and right phylogenies are linear-scaled.

On-device Evolution Trial

Phylometric outcomes. Statistics calculated from reconstruction of 18 million population size/1 million generation on-hardware simulations. Phylometrics were calculated from reconstructions with 10k sampled end-state agents.

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

“

”

Proof-of-concept Experiment

Treatment Conditions. Allowed mutational outcomes under compared outcomes. Note that adaptive regime introduces the possibility for selective sweeps.

a) purifying

regime

b) adaptive

regime

Example reconstructions. Phylogenetic reconstructions of 1 million generation on-hardware simulations. For legibility, phylogeny visualizations were further subsampled to 1k end-state agents. Left phylogenies are log-scaled with ultrametric correction to better show topology and right phylogenies are linear-scaled.

a) purifying regime

b) adaptive regime

👾 =mutate⇒

👾 or 👾^—

👾 =mutate⇒

👾 or 👾^— or 👾⁺

On-device Evolution Trial

Phylometric outcomes. Statistics calculated from reconstruction of 18 million population size/1 million generation on-hardware simulations. Phylometrics were calculated from reconstructions with 10k sampled end-state agents.

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

“

”

simple agent model, explicit fitness

Proof-of-concept Experiment

🧬

Treatment Conditions. Allowed mutational outcomes under compared outcomes. Note that adaptive regime introduces the possibility for selective sweeps.

a) purifying

regime

b) adaptive

regime

Example reconstructions. Phylogenetic reconstructions of 1 million generation on-hardware simulations. For legibility, phylogeny visualizations were further subsampled to 1k end-state agents. Left phylogenies are log-scaled with ultrametric correction to better show topology and right phylogenies are linear-scaled.

a) purifying regime

b) adaptive regime

👾 =mutate⇒

👾 or 👾^—

👾 =mutate⇒

👾 or 👾^— or 👾⁺

On-device Evolution Trial

Phylometric outcomes. Statistics calculated from reconstruction of 18 million population size/1 million generation on-hardware simulations. Phylometrics were calculated from reconstructions with 10k sampled end-state agents.

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

🏝️ 🏝️ 🏝️ 🏝️

“

”

simple agent model, explicit fitness

🧬

🧬

🧬

Example Phylogenies

a) purifying

regime

b) adaptive

regime

Phylometric outcomes. Statistics calculated from reconstruction of 18 million population size/1 million generation on-hardware simulations. Phylometrics were calculated from reconstructions with 10k sampled end-state agents.

📧 morenoma@umich.edu

~ 8 million agents

~ 10k tip phylogenies

a) purifying regime

b) adaptive regime

Phylogeny Structure Metrics

Phylogeny Structure Metrics

📚 [slides, paper, code] https://hopth.ru/de

Phylogeny Structure Metrics

📚 [slides, paper, code] https://hopth.ru/de

Goal: end-to-end encapsulated workflow

raw sim “genome” strings

🧬🧬🧬🧬🧬🧬🧬

🌻

🌸

🌺

🌼

🌻

phylogeny

One Terminal

Command

Goal: end-to-end encapsulated workflow

• raw sim “genome” strings

🧬🧬🧬🧬🧬🧬🧬

2. extract and decode

markers

3. build tree

🌻

🌸

🌺

🌼

🌻

4. phylogeny

courtesy meat-machinery.com

Goal: high-performance, easy workflow

raw sim “genome” strings

🧬🧬🧬🧬🧬🧬🧬

🌻

🌸

🌺

🌼

🌻

phylogeny

1 Terminal

Command

✅ prototype → 🔜 open-source package

Goal: high-performance, easy workflow

• today: 2.5 million tips per minute
• goal: >5 million tips per minute

raw sim “genome” strings

🧬🧬🧬🧬🧬🧬🧬

🌻

🌸

🌺

🌼

🌻

phylogeny

1 Terminal

Command

✅ prototype → 🔜 open-source package

Goal: high-performance, easy workflow

• today: 2.5 million tips per minute
• goal: >5 million tips per minute

raw sim “genome” strings

🧬🧬🧬🧬🧬🧬🧬

🌻

🌸

🌺

🌼

🌻

phylogeny

1 Terminal

Command

✅ prototype → 🔜 open-source package

Goal: high-performance, easy workflow

• today: 2.5 million tips per minute
• goal: >5 million tips per minute
• prototype —> python package

​

Joey

Wagner

​

Connor

Yang

Vivaan Singhvi

raw sim “genome” strings

🧬🧬🧬🧬🧬🧬🧬

🌻

🌸

🌺

🌼

🌻

phylogeny

1 Terminal

Command

✅ prototype → 🔜 open-source package

high-throughput phylogeny generation (in vivo & in silico)

​

contemporary analyses

high-throughput phylogeny generation (in vivo & in silico)

Liu et al., 2024: 235 million seqs

39k jobs, 30.5 hrs

Future Work

Results

agent-based model

beneficial

λ=1 in 100k

deleterious

λ=1 in 1k

poisson distribution

agent-based model

🏝️ 🏝️ 🏝️ 🏝️

​

🏝️ 🏝️ 🏝️ 🏝️

​

🏝️ 🏝️ 🏝️ 🏝️

​

🏝️ 🏝️ 🏝️ 🏝️

1 billion agents

agent-based model

beneficial

λ=1 in 100k

deleterious

λ=1 in 1k

∞

vs

agent-based model

beneficial

λ=1 in 100k

deleterious

λ=1 in 1k

poisson distribution

∞

vs

agent-based model

beneficial

λ=1 in 100k

deleterious

λ=1 in 1k

poisson distribution

max n

1 billion agents

1 billion agents

agent-based model

beneficial

λ=1 in 100k

deleterious

λ=1 in 1k

poisson distribution

agent-based model

beneficial

p=1 in 100k

deleterious

λ=1 in 1k

n loci available

n loci

n loci

n loci

100 million agents

Generations

50/50

Generations

50/50

Generations

1 in 100k

denovo

1 in 100k

denovo

n loci

1 in 100k

denovo

n loci

1 in 100k

denovo

n loci

1 in 100k

denovo

n loci

1 in 100k

denovo

n loci

1 in 100k

denovo

n loci

1 in 100k

denovo

Next Steps

Next Steps

VS

osmotic pressure

antibiotics

Callens et al., 2023

Next Steps

VS

osmotic pressure

more adaptive loci

more hypermutators

antibiotics

fewer adaptive loci

fewer hypermutators

Callens et al., 2023

Q: how many beneficial muts per strain?

Next Steps

VS

osmotic pressure

more adaptive loci

more hypermutators

antibiotics

fewer adaptive loci

fewer hypermutators

Callens et al., 2023

Q: how many beneficial muts per strain?

• more sophisticated fitness landscape models
• well-mixed ABM on GPU
• analyze time to fixation in model vs. theory expectation
• create mathematical model for de novo selection curves
• connect to data from Callens et al or LTEE?

Tagged, Event-driven Programming Model

N

E

S

W

“Ramp”

Tagged, Event-driven Programming Model

“Ramp”

N

E

S

W

“Wavelet”

“Wavelet”

Tagged, Event-driven Programming Model

“Ramp”

N

E

S

W

“Wavelet”

“Wavelet”

Queue

…

fn doTask(wav) {

int c = 10;

…

}

Task

[more about WSE programming model in pocket slides]

GraphCore IPU — another AI/ML accelerator

• 1,200 cores per chip
• clustered up to 1,024 chips

📧 morenoma@umich.edu

Part 1: On-hardware Experiment

Scale and Digital Evolution

📧 morenoma@umich.edu

📚 [these slides] https://hopth.ru/ea

LTEE

(Good et al., 2017)

E.

coli

Scale and Digital Evolution

LTEE

(Good et al., 2017)

Avida

(Ofria and Wilke, 2009)

​

📧 morenoma@umich.edu

📚 [these slides] https://hopth.ru/ea

Image credit: Credit: National Institute of Allergy and Infectious Diseases, National Institutes of Health

​

E.

coli

Scale and Digital Evolution

• both ~billion replications/day

LTEE

(Good et al., 2017)

Avida

(Ofria and Wilke, 2009)

​

📧 morenoma@umich.edu

📚 [these slides] https://hopth.ru/ea

Image credit: Credit: National Institute of Allergy and Infectious Diseases, National Institutes of Health

​

E.

coli

Scale and Digital Evolution

• both ~billion replications/day

​

Cross-scale Phenomena:

• ecological communities
• multicellularity/major transitions

LTEE

(Good et al., 2017)

Avida

(Ofria and Wilke, 2009)

​

📧 morenoma@umich.edu

📚 [these slides] https://hopth.ru/ea

E.

coli

GPU

ALIEN Project (Heinemann, 2024)

MFM/ULAM/T2 Tiles (Ackley, 2023)

Illuninato x Machina

(Ackley, 2010)

📚 [slides, paper, code] https://hopth.ru/de