Inventory 2022: The "inflection point" of life science research - AI intelligence, has it really appeared?

In 2021, artificial intelligence made headlines: Nature, Science published separate articles reporting AI models that can accurately predict protein structure, and at the end of the year, Science also rated AlphaFold as the first
breakthrough of the year 。 In 2022, various artificial intelligence AI technologies followed: DeepMind plans to publish structure predictions totaling more than 200 million, which is almost half the number of all known proteins; For the first time, scientists introduced meta-learning methods in the field of artificial intelligence into the fields of neuroscience and medicine.
Artificial intelligence helps time-resolved cryo-EM to discover the kinetic regulation mechanism of major drug targets; Using AI to identify cancer cells, identify patients with heart complications, classify breast cancer.
.
.
Some scientists even pointed out that artificial intelligence is the "inflection point" of life science research, and artificial intelligence will be able to replace a variety of heavy repetitive experiments in the future, bringing a new era
of scientific research.

This sounds like the plot of science fiction, but it is an undeniable fact that the role of artificial intelligence is increasingly infiltrating life science research, and it is an undeniable fact that the impressive AI helps scientific research in 2022:

What's next for AlphaFold and the AI protein folding revolution?

In July 2021, London-based DeepMind, part of Google's parent company Alphabet, unveiled an artificial intelligence (AI) tool
called AlphaFold2.
The software can predict the three-dimensional structure of proteins from their genetic sequences, and the results are precise
in most cases.

In 2022, DeepMind plans to publish structural projections
totaling more than 200 million.
That's almost half the number of all known proteins: hundreds of times
more than the number of proteins experimentally determined in the Protein Database (PDB) structure library.

AlphaFold also deploys deep learning neural networks: computational architectures inspired by the brain's neural circuitry to discern the kind of
data.
It has been trained
on hundreds of thousands of experimentally determined protein structures and sequences in PDBs and other databases.

Meanwhile, another tech giant is filling dark matter
in the protein universe.
Researchers at Meta (formerly Facebook) used artificial intelligence to predict the structure of about 600 million proteins from bacteria, viruses and other microbes
that have not yet been characterized.
The study was published Nov.
1 on the preprint site BioRxiv
.

In total, the Meta team predicted the structure of more than 617 million proteins, and the work took just two weeks
.
Of those 617 million predictions, the model considered more than 1/3 to be of high quality, so researchers can be confident that the overall shape of the protein is correct, and in some cases, the model can identify finer atomic-level details
.
It is worth mentioning that millions of these structures are completely new, different from
the database of experimentally determined protein structures, or the structure of the AlphaFold database predicted from known organisms.

In some cases, AI saves scientists time; In other cases, it enables previously unimaginable or extremely unrealistic research
.
Despite its limitations, some scientists find its predictions too unreliable
for their work.
However, its rise and the progress of experiments have become unstoppable
.

What's next for AlphaFold and the AI protein folding revolution?

Deep Mind ran into rivals, with Meta AI predicting 600 million protein structures

Meta-learning methods in the field of artificial intelligence are introduced into the fields of neuroscience and medicine

In May, a technical achievement was published
in Nature Neuroscience, a top journal in neurobiology.
This study is the first to introduce meta-learning methods in the field of artificial intelligence into the fields of neuroscience and medicine, which can train reliable AI models on limited medical data and improve the effect
of precision medicine based on brain imaging.

The new method, which has been tested on datasets from the UK Biobank and Human Connectome programme, shows that the new method shows higher accuracy than traditional methods
.
Experiments show that this new training framework is very flexible and can be combined with any machine learning algorithm to effectively train AI predictive models
with good generalization performance on small-scale datasets.

Scientists are bringing AI meta-learning to neuroscience for the first time

Predicting patients at high risk of diabetes complications

Researchers at the University of Houston's Tillman J.
Feltita School of Medicine are developing a clinical decision support system that uses deep learning to predict which patients are more likely to develop complications
.

AI systems predict patients at high risk of diabetes complications

Research shows that artificial intelligence may improve diabetes diagnosis 

Scientists have developed a method that uses artificial intelligence to predict the onset of diabetes within 12 hours

Accurate classification of breast density

A new study shows that an artificial intelligence (AI) tool can accurately and consistently classify
breast density on mammograms.

AI provides accurate breast density classification

A novel protein-small molecule scoring method based on AI

The researchers propose a new method
for constructing unbiased datasets for machine learning scoring function training and testing.
This method introduces four techniques to eliminate hidden biases, given the active molecules of a specific target, based on conditional molecular generation and molecular docking, the corresponding negative samples (decoys) can be efficiently generated based on known active molecules, which provides an unbiased dataset
for the training and evaluation of machine learning scoring functions.

Journal of Medicinal Chemistry: A novel protein-small molecule scoring method based on AI

Artificial intelligence helps time-resolved cryo-EM to discover the kinetic regulation mechanism of major drug targets

In April this year, the team of Mao Youdong of the National Center for Biomedical Imaging Science of Peking University, the Peking University-Tsinghua Joint Center for Life Sciences, and the Center for Quantitative Biology of Peking University published a research paper in the journal Nature, a top international academic journal, reporting the breakthrough scientific discovery
of using self-developed deep learning high-precision four-dimensional reconstruction technology to develop and apply time-resolved cryo-EM to elucidate the mechanism of kinetic regulation and conformational reprogramming of human proteasomes at the atomic level 。 This is the first time in the world that artificial intelligence four-dimensional reconstruction technology is used to greatly improve the accuracy of time-resolved cryo-EM analysis, realize the world's leading original first-class results in atomic-level kinetic observation of major disease target complexes, and demonstrate a new class of protein composite dynamics research paradigm
.

Nature of Peking University published a breakthrough research result: artificial intelligence helps time-resolved cryo-EM to discover the kinetic regulation mechanism of major drug targets

Detect early signs of pancreatic cancer

An artificial intelligence (AI) tool developed by Cedars Sinai researchers accurately predicted who would develop pancreatic cancer based on images from CT scans they had in the years before they were diagnosed with pancreatic cancer
.

Artificial intelligence can detect early signs of pancreatic cancer

A new method for early detection of lung cancer combined with artificial intelligence combined with metabolome 

Artificial intelligence and machine learning show promise in cancer diagnosis and treatment

Genome Biology: Artificial intelligence identifies cancer cells

A new artificial intelligence blood test can detect liver cancer

Heart disease

Scientists at the University of Utah School of Health have demonstrated for the first time that artificial intelligence can better predict the onset and course
of cardiovascular disease.
Working with doctors at Intermountain Junior Children's Hospital, the researchers developed a unique computational tool to precisely measure the synergistic effects
of existing medical conditions on the heart and blood vessels.

PLOS Digital Health: AI identifies individuals at risk of heart disease complications

Promoting heart health through artificial intelligence

New AI tools can detect heart disease that is often overlooked 

Eye diseases

Scientists from the National Eye Institute (NEI) have identified five subpopulations of the retinal pigment epithelium (RPE), a layer of tissue
that nourishes and supports retinal photoreceptor cells.
Using artificial intelligence, the researchers analyzed RPE images at single-cell resolution to create a reference map that locates each subpopulation
within the eye.
The report of the study was published May 6, 2022 in
the Proceedings of the National Academy of Sciences.

Treating blinding eye diseases with artificial intelligence 

More diseases

Using artificial intelligence to improve TB treatment

Artificial intelligence enables early, non-invasive, accurate screening for Down syndrome

Artificial intelligence helps detect gait changes and diagnose Parkinson's disease