Machine learning spots molecules that work even against ‘untreatable’ strains of bacteria.

by Jo Marchant

A pioneering machine-learning approach has identified powerful new types of antibiotic from a pool of more than 100 million molecules — including one that works against a wide range of bacteria, including tuberculosis and strains considered untreatable.

The researchers say the antibiotic, called halicin, is the first discovered with artificial intelligence (AI). Although AI has been used to aid parts of the antibiotic-discovery process before, they say that this is the first time it has identified completely new kinds of antibiotic from scratch, without using any previous human assumptions. The work, led by synthetic biologist Jim Collins at the Massachusetts Institute of Technology in Cambridge, is published in Cell1.

The study is remarkable, says Jacob Durrant, a computational biologist at the University of Pittsburgh, Pennsylvania. The team didn’t just identify candidates, but also validated promising molecules in animal tests, he says. What’s more, the approach could also be applied to other types of drug, such as those used to treat cancer or neurodegenerative diseases, says Durrant.

Bacterial resistance to antibiotics is rising dramatically worldwide, and researchers predict that unless new drugs are developed urgently, resistant infections could kill ten million people per year by 2050. But over the past few decades, the discovery and regulatory approval of new antibiotics has slowed. “People keep finding the same molecules over and over,” says Collins. “We need novel chemistries with novel mechanisms of action.”

Forget your assumptions
Collins and his team developed a neural network — an AI algorithm inspired by the brain’s architecture — that learns the properties of molecules atom by atom.

The researchers trained its neural network to spot molecules that inhibit the growth of the bacterium Escherichia coli, using a collection of 2,335 molecules for which the antibacterial activity was known. This includes a library of about 300 approved antibiotics, as well as 800 natural products from plant, animal and microbial sources.

The algorithm learns to predict molecular function without any assumptions about how drugs work and without chemical groups being labelled, says Regina Barzilay, an AI researcher at MIT and a co-author of the study. “As a result, the model can learn new patterns unknown to human experts.”

Once the model was trained, the researchers used it to screen a library called the Drug Repurposing Hub, which contains around 6,000 molecules under investigation for human diseases. They asked it to predict which would be effective against E. coli, and to show them only molecules that look different from conventional antibiotics.

From the resulting hits, the researchers selected about 100 candidates for physical testing. One of these — a molecule being investigated as a diabetes treatment — turned out to be a potent antibiotic, which they called halicin after HAL, the intelligent computer in the film 2001: A Space Odyssey. In tests in mice, this molecule was active against a wide spectrum of pathogens, including a strain of Clostridioides difficile and one of Acinetobacter baumannii that is ‘pan-resistant’ and against which new antibiotics are urgently required.

Proton block
Antibiotics work through a range of mechanisms, such as blocking the enzymes involved in cell-wall biosynthesis, DNA repair or protein synthesis. But halicin’s mechanism is unconventional: it disrupts the flow of protons across a cell membrane. In initial animal tests, it also seemed to have low toxicity and be robust against resistance. In experiments, resistance to other antibiotic compounds typically arises within a day or two, says Collins. “But even after 30 days of such testing we didn’t see any resistance against halicin.”

The team then screened more than 107 million molecular structures in a database called ZINC15. From a shortlist of 23, physical tests identified 8 with antibacterial activity. Two of these had potent activity against a broad range of pathogens, and could overcome even antibiotic-resistant strains of E. coli.

The study is “a great example of the growing body of work using computational methods to discover and predict properties of potential drugs”, says Bob Murphy, a computational biologist at Carnegie Mellon University in Pittsburgh. He notes that AI methods have previously been developed to mine huge databases of genes and metabolites to identify molecule types that could include new antibiotics2,3.

But Collins and his team say that their approach is different — rather than search for specific structures or molecular classes, they’re training their network to look for molecules with a particular activity. The team is now hoping to partner with an outside group or company to get halicin into clinical trials. It also wants to broaden the approach to find more new antibiotics, and design molecules from scratch. Barzilay says their latest work is a proof of concept. “This study puts it all together and demonstrates what it can do.”

doi: 10.1038/d41586-020-00018-3
Stokes, J. M. et al. Cell (2020).

Police are investigating after jars containing preserved human tongues were found in the crawlspace of a house in northwest Gainesville home, WCJB-TV is reporting.

Gainesville police said the remains were discovered during an inspection of the foundation of a home that was once owned by Dr. Ronald Baughman, a former University of Florida researcher and current professor emeritus who published studies in the 1970s and 80s.

Some of the jars date back as far as the 1960s.

Police told the station they were looking into the possibility that the preserved tongues are related to work that Baughman may have brought home and stored under the house’s floorboards.

Shorter sleep duration among children was associated with increased risk for depression, anxiety, impulsive behavior and poor cognitive performance, according to study findings published in Molecular Psychiatry.

“Sleep disturbances are common among children and adolescents around the world, with approximately 60% of adolescents in the United States receiving less than 8 hours of sleep on school nights,” Jianfeng Feng, PhD, of the department of computer science at University of Warwick in the UK, told Healio Psychiatry. “An important public health implication is that psychopathology in both children and their parents should be considered in relation to sleep problems in children. Further, we showed that brain structure is associated with sleep problems in children and that this is related to whether the child has depressive problems.”

According to Feng and colleagues, the present study is the first large-scale research effort to analyze sleep duration in children and its impact on psychiatric problems including depression, brain structure and cognition. They analyzed measures related to these areas using data from the Adolescent Brain Cognitive Development Study, which included structural MRI data from 11,067 individuals aged 9 to 11 years.

The researchers found that depression, anxiety and impulsive behavior were negatively correlated with sleep duration. Dimensional psychopathology in participants’ parents was correlated with short sleep duration in the children. Feng and colleagues noted that the orbitofrontal cortex, prefrontal and temporal cortex, precuneus and supramarginal gyrus were brain areas in which higher volume was correlated with longer sleep duration. According to longitudinal data analysis, psychiatric problems, particularly depressive problems, were significantly associated with short sleep duration 1 year later. Moreover, they found that depressive problems significantly mediated these brain regions’ effect on sleep. Higher volume of the prefrontal cortex, temporal cortex and medial orbitofrontal cortex were associated with higher cognitive scores.

“Our findings showed that 53% of children received less than 9 hours of sleep per night,” Feng said. “More importantly, the behavior problems total score for children with less than 7 hours of sleep was 53% higher on average and the cognitive total score was 7.8% lower on average than for children with 9 to 11 hours of sleep. We hope this study attracts public attention to sleep problems in children and provides evidence for governments to develop advice about sleep for children.” – by Joe Gramigna

A promising molecule has offered hope for a new treatment that could stop or slow Parkinson’s, something no treatment can currently do.

Researchers from the University of Helsinki found that molecule BT13 has the potential to both boost levels of dopamine, the chemical that is lost in Parkinson’s, as well as protect the dopamine-producing brain cells from dying.

The results from the study, co-funded by Parkinson’s UK and published online today in the journal Movement Disorders, showed an increase in dopamine levels in the brains of mice following the injection of the molecule. BT13 also activated a specific receptor in the mouse brains to protect the cells.

Typically, by the time people are diagnosed with Parkinson’s, they have already lost 70-80 per cent of their dopamine-producing cells, which are involved in coordinating movement.

While current treatments mask the symptoms, there is nothing that can slow down its progression or prevent more brain cells from being lost, and as dopamine levels continue to fall, symptoms get worse and new symptoms can appear.

Researchers are now working on improving the properties of BT13 to make it more effective as a potential treatment which, if successful, could benefit the 145,000 people living with Parkinson’s in the UK.

The study builds on previous research on another molecule that targets the same receptors in the brain, glial cell line-derived neurotrophic factor (GDNF), an experimental treatment for Parkinson’s which was the subject of a BBC documentary in February 2019. While the results were not clear cut, GDNF has shown promise to restore damaged cells in Parkinson’s.

However, the GDNF protein requires complex surgery to deliver the treatment to the brain because it’s a large molecule that cannot cross the blood-brain barrier – a protective barrier that prevents some drugs from getting into the brain.

BT13, a smaller molecule, is able to cross the blood-brain barrier – and therefore could be more easily administered as a treatment, if shown to be beneficial in further clinical trials.

Professor David Dexter, Deputy Director of Research at Parkinson’s UK, said:

“People with Parkinson’s desperately need a new treatment that can stop the condition in its tracks, instead of just masking the symptoms.

“One of the biggest challenges for Parkinson’s research is how to get drugs past the blood-brain barrier, so the exciting discovery of BT13 has opened up a new avenue for research to explore, and the molecule holds great promise as a way to slow or stop Parkinson’s.

“More research is needed to turn BT13 into a treatment to be tested in clinical trials, to see if it really could transform the lives of people living with Parkinson’s.”

Dr Yulia Sidorova, lead researcher on the study, said: “We are constantly working on improving the effectiveness of BT13. We are now testing a series of similar BT13 compounds, which were predicted by a computer program to have even better characteristics.

“Our ultimate goal is to progress these compounds to clinical trials in a few coming years.”

Molecule offers hope for halting Parkinson’s

Dr. Anjali Rajadhyaksha
Professor of Neuroscience in Pediatrics
Associate Dean of Program Development
Weill Cornell Graduate School

Dr. Francis Lee
Psychiatry/Pharmacology; Chair and Psychiatrist-in-Chief
Mortimer D. Sackler, M.D. Professor in Psychiatry, Weill Cornell Medicine

Dr. Caitlin Burgdorf

A common variation in a human gene that affects the brain’s reward processing circuit increases vulnerability to the rewarding effects of the main psychoactive ingredient of cannabis in adolescent females, but not males, according to preclinical research by Weill Cornell Medicine investigators. As adolescence represents a highly sensitive period of brain development with the highest risk for initiating cannabis use, these findings in mice have important implications for understanding the influence of genetics on cannabis dependence in humans.

The brain’s endocannabinoid system regulates activity of cannabinoids that are normally produced by the body to influence brain development and regulate mood, as well as those from external sources, such as the psychoactive ingredient THC, also known as Δ9-tetrahydrocannabinol, which is found in cannabis. An enzyme called fatty acid amide hydrolase (FAAH) breaks down a cannabinoid called anandamide that is naturally found in the brain and is most closely related to THC, helping to remove it from circulation.

In the study, published Feb. 12 in Science Advances, the investigators examined mice harboring a human gene variant that causes FAAH to degrade more easily, increasing overall anandamide levels in the brain. They discovered that the variant resulted in an overactive reward circuit in female—but not male adolescent mice—that resulted in higher preference for THC in females. Previous clinical studies linked this FAAH variant with increased risk for problem drug use, but no studies had specifically looked at the mechanistic effect on cannabis dependence.

“Our study shows that a variant in the FAAH gene, which is found in about one-third of people, increases vulnerability to THC in females and has large-scale impact on brain regions and pathways responsible for processing reward,” said lead author Dr. Caitlin Burgdorf, a recent doctoral graduate from the Weill Cornell Graduate School of Medical Sciences. “Our findings suggest that genetics can be a contributing factor for increased susceptibility to cannabis dependence in select populations.”

The team found that female mice with the FAAH variant showed an increased preference for the environment in which they’d been exposed to THC over a neutral environment when they were exposed to the substance during adolescence, and the effect persisted into adulthood. However, if female mice with this variant were exposed to THC for the first time in adulthood, there was no increased preference for THC. These findings in mice parallel observations in humans that a select population of females are more sensitive to the effects of cannabis and demonstrate a quicker progression to cannabis dependence. “Our findings suggest that we have discovered a genetic factor to potentially identify subjects at risk for cannabis dependence,” said Dr. Burgdorf.

The investigators also found that the genetic variant led to increased neuronal connections and neural activity between two regions of the brain heavily implicated in reward behavior. Next, the team reversed the overactive reward circuit in female mice and found that decreasing circuit activity dampened the rewarding effects of THC.

As substance abuse disorders often emerge during adolescence, the investigators say this study has significant implications for translating these findings to inform developmental and genetic risk factors for human cannabis dependence.

“Our study provides new insights into cannabis dependence and provides us with a circuit and molecular framework to further explore the mechanisms of cannabis dependence,” said co-senior author Dr. Anjali Rajadhyaksha, professor of neuroscience in pediatrics and associate professor of neuroscience in the Feil Family Brain and Mind Research Institute and a member of the Drukier Institute for Children’s Health at Weill Cornell Medicine.

Although genetic factors are increasingly found to be associated with risk for other types of addiction, very few studies have investigated genetic factors associated with increasing risk for cannabis dependence. “In the future, we could use the presence of this FAAH genetic variant to potentially predict if an individual is more likely to be vulnerable to cannabis dependence,” said co-senior author, Dr. Francis Lee, chair of the Department of Psychiatry at Weill Cornell Medicine and psychiatrist-in-chief at NewYork-Presbyterian/Weill Cornell Medical Center. “We are getting one step closer to understanding exactly how neurodevelopmental and genetic factors play interrelated roles to increase susceptibility for cannabis dependence.”

Additional authors on the study were Dr. Deqiang Jing, Ruirong Yang and Chienchum Huang from the Department of Psychiatry at Weill Cornell Medicine; Drs. Teresa A. Milner and Dr. Virginia M. Pickel from the Feil Family Brain and Mind Research Institute at Weill Cornell Medicine; Dr. Matthew N. Hill from departments of Cell Biology and Anatomy and Psychiatry at University of Calgary; and Dr. Ken Mackie from the Department of Psychological and Brain Sciences at Indiana University Bloomington.

This research was supported by the National Institute of Health (Grants T32DA039080, R01DA08259, R01HL098351, R01HL136520, R01DA042943, R01NS052819, R01DA029122), Weill Cornell’s Mowrer Memorial Graduate Student Fellowship, NewYork-Presbyterian Youth Anxiety Center, the Pritzker Neuropsychiatric Disorders Research Consortium, the DeWitt-Wallace Fund of the New York Community Trust, and The Paul Fund.

The Harvard Medical School researcher’s work on the genetic basis of protein coding and production led him to make groundbreaking discoveries in immunology, molecular biology, and cancer genetics.


Harvard Medical School molecular geneticist Philip Leder died last week (February 2). He was 85.

Leder was revered for his work in molecular biology, immunology, and cancer genetics. His first scientific breakthrough came in the 1960s when he was working as a postdoc in geneticist Marshall Nirenberg’s lab at the National Institutes of Health (NIH). Together they developed a technique that confirmed that amino acids were encoded by a sequence of three nucleotides and revealed the triplet code of ambiguous amino acids.

From there, Leder went on to determine the first complete sequence of a mammalian gene, develop the first recombinant DNA vector system safe for use in the lab, identify the structure of genes that encode antibody molecules, discover a gene that caused cancer, and develop the first mouse model of cancer.

“Phil Leder was special. Among great scientists, he was special, and among scientists, he was an icon,” David Livingston, a geneticist at Harvard who worked in Leder’s lab at NIH, tells The Scientist. “He was gifted. He was generous. He was a splendid person to listen to talk, to run experiments by, and be criticized by. He was a splendid human being on top of all of it.”

Leder was born on November 19, 1934 in Washington, DC, and grew up there. He attended Western High School, graduated in 1952, and went on to study at Harvard University. He interned at NIH as an undergraduate, working in biochemist Martha Vaughan’s lab in the National Heart Institute, which is now the National Heart, Lung, and Blood Institute. He finished his bachelor’s degree at Harvard in 1956 and stayed there for medical school, graduating in 1960.

After a two-year residency program at the University of Minnesota Hospitals, he returned to NIH to work with Nirenberg. Leder dove headfirst into the race to decipher the way genes encode proteins and helped to design a filtering instrument to rapidly test 45 amino acid samples simultaneously, instead of one at a time. Leder and Nirenberg could quickly tag amino acids with a radioactive label, bind them to triplet RNA sequences, and put them into the filtering instrument, which helped the team decode unknown amino acid codon sequences, well before other scientists could, according to a remembrance on Leder posted by NIH.

It was one of the most exciting times in Leder’s life, he said. “I would go to bed thinking about the next day’s experiments and then jump out of bed in the morning and rush to the laboratory,” he recalled in a 2012 interview with American Society for Biochemistry and Molecular Biology Today. “I stayed late at night. It was a lot of work, but the intellectual excitement was enormous.” The two published their work on the codons in 1964.

Leder’s “work w/Marshall Nirenberg set the stage for the revolution in molecular genetics,” NIH director Francis Collins wrote on Twitter last Friday (February 7).

In 1965, Leder joined the Weizmann Institute in Rehovot, Israel, as a visiting scientist and stayed until 1966. He returned to the NIH, serving as a research medical officer in the National Cancer Institute from 1966 to 1969 and then became head of the Section on Molecular Genetics in the Laboratory of Molecular Genetics in the National Institute of Child Health and Human Development and in 1972 was promoted to the director of the lab.

During this time and through the 1970s, he and his colleagues worked on deciphering the genetic sequence of alpha globin, a component of hemoglobin, a protein in red blood cells that carries oxygen to the body’s cells and tissues. His work also revealed important details about the genetics of encoding antibodies and that the synthesis of antibodies was not only regulated by genetics but also biochemical processes that ensure specificity to target the right antigen presented by viruses, bacteria, or other invaders in the body.

What made Leder such an outstanding scientist, Livingston explains, was his immense rigor. Control experiments, for example, had to be “at least as incisive or demanding and rigorous as the actual experiments . . . to prove that nothing in the discovery experiment was an artifact,” he says. “And he had an immensely adventurous mind. No problem was beyond at least discussion,” which made Leder unique as a mentor. “In fact, his ability to mentor was internationally celebrated,” Livingston explains. “You could listen to his talks, and you knew he was a fantastic teacher because his mind was utterly clear.”

Leder joined Harvard Medical School (HMS) in 1980, founding its genetics department in 1981 and chairing the department for 25 years. His research there led to the discovery of a specific gene, MYC. With Harvard colleague Timothy Stewart, Leder began using a fine glass needle to insert the cancer-causing gene into mouse embryos just after fertilization, thereby creating OncoMouse, a genetic line of mice that were prone to developing the disease. The duo patented the animal in 1988, giving researchers an unprecedented tool to study cancer and how to treat it.

His work at Harvard was not limited to his research. He made fundamental changes to hiring, instituting nationwide searches for new assistant professors in the genetics department, which increased the likelihood of hiring women, notes Jonathan Seidman, a geneticist at Harvard who worked in Leder’s lab at NIH in the 1970s. Leder also made sure the department didn’t get too big, Seidman says, and he insisted that if faculty were on different floors, spiral staircases—rather than drab stairwells—would connect them, making it easy for researchers to communicate and collaborate.

Leder’s “contributions to science and to HMS cannot be overstated, and he will never be forgotten,” George Daley, Harvard’s dean of the faculty of medicine wrote to colleagues on February 4.

For his work, Leder was honored with the Albert Lasker Award for Basic Medical Research, the US National Medal of Science, the Heineken Prize from the Royal Netherlands Academy of Arts of Sciences, and the William Allan Medal from the American Society of Human Genetics. He was a member of the National Academy of Sciences, a Fellow of the American Association for the Advancement of Science, and a Howard Hughes Medical Institute investigator.

Surviving him are his wife, Aya Leder, his children, Micki, Tani, and Ben, his daughters-in-law, Karen Leder and Mary Leder, and his grandchildren, Jacob, David, Sarah, Eli, Alex, Matt, Amanda, and Annie.

Ashley Yeager is an associate editor at The Scientist. Email her at Follow her on Twitter @AshleyJYeager.–who-deciphered-amino-acid-sequences–dies-67096

Researchers from Case Western Reserve University School of Medicine, University Hospitals Cleveland Medical Center (UH), Cleveland Clinic and Lifebanc (a Northeast Ohio organ-procurement organization) have developed a new way to preserve donated kidneys–a method that could extend the number and quality of kidneys available for transplant, saving more people with end-stage renal disease, more commonly known as “kidney failure.”

The team identified a drug–ethyl nitrite–that could be added to the preservation fluid to generate tiny molecules called S-nitrosothiols (SNOs), which regulate tissue-oxygen delivery. This, in turn, restored flow-through and reduced resistance within the kidney. Higher flow-rates and lower resistance are associated with better kidney function after transplantation.

Their research was funded by a grant from the Roche Organ Transplant Research Foundation and recently published in Annals of Surgery.

The United States has one of the world’s highest incidences of end-stage renal disease, and the number of afflicted individuals continues to increase. The prevalence of end-stage renal disease has more than doubled between 1990 and 2016, according to the Centers for Disease Control.

The optimal treatment is a kidney transplant, but demand far exceeds supply. Additionally, donation rates for deceased donors have been static for several years, despite various public-education campaigns, resulting in fewer kidneys available for transplant. And while the proportion and number of living donors has increased, this latter group still only makes up a small percentage of recovered kidneys for transplant.

Increasing the number of kidneys available for transplant benefits patients by extending lifespans and/or enhancing quality of life as well as the potential for reducing medical costs (a transplant is cheaper than ongoing dialysis). To help improve outcomes for kidney transplant patients, the team explored ways to extend the viability of donated kidneys.

Improvements in surgical techniques and immunosuppression therapies have made kidney transplants a relatively common procedure. However, less attention has been paid to maintaining/improving kidney function during the kidney-transport phase.

“We addressed this latter point through developing enhanced preservation methods,” said senior author James Reynolds, professor of Anesthesiology and Perioperative Medicine at Case Western Reserve School of Medicine and a member of the Harrington Discovery Institute at UH.

For decades, procured kidneys were simply flushed with preservation solution and then transported in ice-filled coolers to the recipient’s hospital. But advances in pumping technology slowly changed the field toward active storage, the preferred method for conveying the organ from donor to recipient.

“However, while 85% of kidneys are now pumped, up to 20% of kidneys are determined to be unsuitable for transplant during the storage phase,” said Kenneth Chavin, professor of surgery at the School of Medicine, chief of hepatobiliary and transplant surgery and director of the UH Transplant Institute.

“For several years, our team has directed research efforts toward understanding and improving the body’s response to medical manipulation,” Reynolds said. “Organ-donor physiology and ‘transport status’ fit well within this metric. We identified a therapy that might improve kidney perfusion, a significant factor in predicting how the organ will perform post-transplant.”

Previous work by Reynolds and long-time collaborator Jonathan Stamler, the Robert S. and Sylvia K. Reitman Family Foundation Distinguished Chair in Cardiovascular Innovation and president of the Harrington Discovery Institute, determined that brain death significantly reduces SNOs, which impairs blood-flow and tissue-oxygenation to the kidneys and other commonly transplanted organs. The loss of SNOs is not corrected by current preservation fluids, so impaired flow through the kidneys continues during storage and transport.