7

NEWS

June 2024 ESRFnews

Deep learning boost to

additive manufacturing

Images from the ESRF’s ID19 beamline

have helped in the training of a new

deep-learning model to assist in the

X-ray analysis of additive manufacturing

processes. Known as AM-SegNet, the

model allows scientists to automatically

process large radiographic datasets,

thereby allowing deeper insights into

complex interactions during laser-

based, 3D printing of metals.

Imaging technologies are crucial in

understanding why problems occur

during additive manufacturing, as they

help to identify defects at various stages

of the process and ways to improve the

quality of components. Particularly in

the case of synchrotron X-ray imaging,

however, there are often large volumes

of imagery, making manual data-

processing very time-consuming.

Now researchers from University

College London, UK, have built

an accurate and efficient deep-

learning model to perform automatic

segmentation and quantification,

trained with over 10,000 images

from three different synchrotron

facilities, including the ESRF at the

ID19 beamline. It works by assigning

a specific label to each pixel within

an image, allowing for feature

quantification and correlation across

a large dataset with high confidence.

In tests, it proved to have the highest

segmentation accuracy (

~

96%) as well

as the fastest processing speed (less

than 4 ms per frame outperforming

other stateoftheart models Virtual

Phys Prototyp e2325572

The researchers believe that the

models automatic analysis has

potential applications for many other

advanced manufacturing processes

This development anticipates a near

future where highspeed synchrotron

experiments will have their images

segmented and quantified in real

time through deep learning the

researchers say

V U E D I C I.O R G

BM29 identif ies fractal molecule

Scientists from Germany have used the

ESRF’s X-rays to help discover the first

fractal molecule ever found in nature.

The microbial enzyme spontaneously

assembles into a pattern known as the

Sierpinski triangle.

From jagged coastlines to

romanesco broccoli, fractals are found

all over the natural world. Their

mesmerising shapes defy conventional

geometry, and possess infinite detail.

The discovery now of the first

fractal at the molecular scale has

been made by a group led by the

Max Planck Institute and Philipps

University in Marburg, and came

about via the study of a microbial

enzyme, citrate synthase, which is

found in cyanobacteria. Normally

citrate synthases form relatively small

and symmetrical assemblies, but when

the researchers studied theirs with

mass photometry, they realised that it

was unusually big.

A subsequent study with cryo-

electron microscopy (cryo-EM)

revealed that the molecule had a

pattern resembling a Sierpinski

triangle, a famous type of fractal.

However, it took small-angle X-ray

scattering at the ESRF’s BM29

beamline to show that the

assemblies could replicate over

several levels in true fractal

fashion (Nature doi:10.1038/

s41586-024-07287-2).

“The ESRF was very important,

because it helped us show that

these assemblies can keep growing

infinitely,” says Georg Hochberg,

an evolutionary biologist at the Max

Planck Institute.

An enzyme, citrate

synthase, from a

cyanobacterium is

the first fractal

molecule ever

recorded.

M P I F. T E R R E S T R I A L M I C R O B I O L O G Y/ H O C H B E R G-

Hallucinogen synthesis revealed

ESRF users have uncovered the process

behind the biosynthesis of psilocybin,

the natural hallucinogen found in

magic mushrooms. The discovery

could lead to tailor-made variants

for the treatment of mental health

conditions such as severe depression

The users led by a group at the

Medical University of Innsbruck in

Austria relied on the ESRF structural

biology beamlines to help determine

the structure of a crucial enzyme in

psilocybin biosynthesis at various stages

of the reaction cycle The enzyme

PsiM is responsible for the last step

in psilocybin synthesis whereby two

methyl groups are added to a specific

molecule called norbaeocystin

The results showed that PsiM is

incredibly specific wrapping around

norbaeocystin tightly while guiding

it through a series of precise chemical

reactions (Nat. Commun. 15 2709).

There is a growing interest in the

potential medical uses of psilocybin,

with efforts to bioengineer the

compound and investigate the

enhanced therapeutic properties of

new analogues The latest findings will

help to guide this type of research

Using the powerful ID231

beamline at the ESRF helped us to

obtain the best possible data and

reach the exceptional final resolution

of 09 Å says Sebastiaan Werten a

structural biologist at the Medical

University of Innsbruck This is

an astonishing level of detail which

enables you to observe individual

atoms within the enzyme

“ID23-1

helped us to

obtain the best

possible data”