NATURE PODCAST

Podcast: China's growing science network, and talking brain signals

Benjamin Thompson and Shamini Bundell bring you the latest science updates.

This week, China’s Belt and Road Initiative, and translating brain patterns into speech.

In this episode:

00:46 China’s Belt and Road

How might China’s enormous infrastructure-building project affect global science? Editorial: Build a sustainable Belt and Road; News Feature: How China is redrawing the map of world science; News Feature: Scientists in Pakistan and Sri Lanka bet their futures on China

07:54 Research Highlights

A surprising snake find, and improving a cancer therapy. Research Highlight: A viper’s tooth in ancient human poo hints at snake-eating rituals; Research Highlight: A cancer therapy’s grave dangers subside after a little fine-tuning

10:20 Making waves in synthesising speech

Researchers train a computer to translate brain patterns into speech. Research article: Anumanchipalli et al.; News and Views: Brain implants that let you speak your mind; News: Brain signals translated into speech using artificial intelligence

19:04 News Chat

An ancient jawbone sheds new light on the mysterious Denisovans, and gender bias in grant reviews. News: Biggest Denisovan fossil yet is the first found outside a Siberian cave; News: Male researchers’ ‘vague’ language more likely to win grants

Never miss an episode: Subscribe to the Nature Podcast on iTunes, Google Podcasts, or your favourite podcast app. Head here for the Nature Podcast RSS feed.

Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.

Transcript

This week, China’s Belt and Road Initiative, and translating brain patterns into speech.

Host: Benjamin Thompson

Welcome back to the Nature Podcast. This week, we’ll be hearing about a giant infrastructure project from China…

Host: Shamini Bundell

And finding out how researchers are translating brain patterns into speech. I’m Shamini Bundell.

Host: Benjamin Thompson

And I’m Benjamin Thompson.

[Jingle]

Interviewer: Benjamin Thompson

Over the next couple of weeks, Nature is publishing a series of feature articles that focus on China’s massive Belt and Road Initiative. This project will connect the country to others around the world by building infrastructure including high-speed rail networks, airports and even whole cities. To date, 126 counties from Africa, Asia, Europe and South America have signed up to the project, which is designed to transform the global movement of goods and services. The features in Nature are looking at the effect that the Belt and Road Initiative is having on scientific research in different parts of the world. The first two of these features are written by science policy journalist Ehsan Masood, who dropped by to tell us more.

Interviewee: Ehsan Masood

So, the Belt and Road is six years old. When it finishes, it will be the largest infrastructure building project coming out of one nation since probably the Marshall Plan in which the United States mostly financed the rebuilding of Europe after the end of World War II. It was a mix of grants and loans to European nations. This is mostly loans and some grants from China to the nations that go towards its west and leading up all the way to eastern Europe.

Interviewer: Benjamin Thompson

Conservative estimates put the cost of the Belt and Road Initiative at over US$1 trillion and the project, which was launched in 2013 by China’s President Xi Jinping, is designed to connect China to countries around the world by an expansive network of land and sea routes.

Interviewee: Ehsan Masood

It’s about building railway lines, it’s about building motorways, it’s about building airports and it’s about either building new ports or Chinese companies taking over the management and the running of seaports.

Interviewer: Benjamin Thompson

The Belt and Road Initiative, or BRI as it’s sometimes known, aims to transform China’s global trade networks and access to markets, but shortly after its launch, it became clear that science was going to be a big part of the initiative as well. This aspect of the Belt and Road is overseen by the Chinese Academy of Sciences President, Bai Chunli, who last year wrote in the Bulletin of the Chinese Academy of Sciences that “Science, technology and innovation are the core driving force for the BRI development.” The Chinese Academy of Sciences has invested over US$250 million into Belt and Road science and technology projects, opening centres of excellence in China and building research and training centres across the world. It’s also setting up the Digital Belt and Road as a data sharing platform for participating countries. Alongside all of this comes a big investment in the training of PhD students from Belt and Road countries.

Interviewee: Ehsan Masood

Just in Pakistan, so in one country, right now the total numbers of scholarship holders, masters and PhD scholars, is about 7,000 per year. Now, just a few weeks ago, the Chinese ambassador to Islamabad said that he wants to increase that number to 20,000, and this is just like one country. So, you’re looking at quite large numbers of people who are right now very young, they’re going to go to China, they’re going to do their PhDs and learn Mandarin, and when they come back they will outnumber the English-speaking community of scientists and PhD holders. So, China is really looking to the long term in both building capacity but also making a sort of very strong case for very China-friendly scientists into the future.

Interviewer: Benjamin Thompson

The majority of these students are studying in applied sciences or engineering, and are being trained in areas of particular importance to their home countries. It seems like a win-win – China builds its global research networks and the countries build relevant research capacity.

Interviewee: Ehsan Masood

Within the countries themselves, and particularly at the level of the early career researcher, there is hardly any criticism of what’s happening here. I think people are genuinely surprised, but in a good way, that a superpower like China is suddenly interested in the training of scientists because America stopped doing this a long time ago on this scale – it did happen, but in the 60s. The European Union countries stopped doing this on this scale a while back and this is why the Chinese contribution is really quite something in that sense. There are of course problems but in this aspect, it’s a really positive thing.

Interviewer: Benjamin Thompson

As part of his reporting, Ehsan visited countries that have signed up to the Belt and Road, to see how the initiative could make a difference to local populations. For example, in Sri Lanka, he saw how local researchers are collaborating with Chinese scientists to investigate the cause of a mysterious kidney disease affecting the rural population. But despite the positivity of researchers involved in projects like these, issues have been raised about the broader Belt and Road Initiative as infrastructure projects take off around the world.

Interviewee: Ehsan Masood

So, there’s a few concerns – one is transparency. It’s very difficult to get hard numbers on anything unless you know people to know where the projects are, what’s happening. There isn’t a sort of central place where you can find this out. Environmental impact is another quite serious concern, particularly in Pakistan, but it other countries as well. Most of the energy projects are coal-fired power stations in that country and it’s going to have huge environmental impacts.

Another issue is debt, though this relates less to the science because the funding that’s coming in is not Chinese loans but grants. They’ll be match funded or co-funded by the countries, but on other Belt and Road projects it’s almost entirely loans. A lot of the countries are going to find themselves, and some of them already have found themselves, in quite serious debt. We’re looking at somewhere between sort of 8, 10, 12% of a country’s total debt that it owes to others is just from one country – China. That’s quite a serious amount of exposure and we don’t really know going forward where that might lead.

Interviewer: Benjamin Thompson

And these concerns don’t appear to have gone unnoticed. Last week, China hosted the second Belt and Road forum, attended by heads of state and representatives from governments around the world. At the event, President Xi mentioned the need to ensure that Belt and Road projects are done in a transparent way, and that they should be financially sustainable. He also said that that the initiative needs to be clean and green. It’s now six years since President Xi announced the start of the Belt and Road Initiative, but it’s too early to tell how this enormous project might affect scientific research around the world, and whether it will indeed be a win-win for everybody involved. What is clear though is that China is in this for the long haul, and we can expect to hear a lot more about Belt and Road in the years to come. In the meantime, there’s a more in-depth discussion of the project and the questions surrounding it over at nature.com/news, where you’ll find Ehsan’s features and an editorial.

Host: Shamini Bundell

Later in the show, we’ll be hearing the latest about the mysterious ancient hominin group known as the Denisovans – that’s coming up in the News Chat. Now though, Anna Nagle is here with this week’s Research Highlights.

[Jingle]

Anna Nagle

Sifting through ancient human faeces may not be your dream job, so you at least hope that someone doing it comes away with something interesting. So, how about signs of ritualistic snake eating? A team in the US analysed a 1,500-year-old sample of human dung found in Texas. In it they found bones, scales and a 1-centimetre-long fang from a venomous snake. Now, the ancient peoples of North American are known to have eaten snakes, even venomous ones, but they usually removed the head and the scales. The fact that the team found these parts in the faecal remains suggests the reptile was eaten whole and unskinned. Because many ancient peoples worshipped snakes, the team suggests that this slithery specimen was therefore likely eaten as part of a ritualistic event. They did also find the remains of a small rodent, again eaten whole and raw, but it’s not clear if that was the human’s last meal or the snake’s. Digest that research over at the Journal of Archaeological Science Reports.

[Jingle]

Anna Nagle

Tweaking T-cells has eliminated some of the serious side effects associated with a promising cancer treatment. CAR T therapy involved genetically engineering a patient’s T cells – a type of immune cell – and using them to fight the patient’s cancer. But the treatment can have serious side effects, including multi-organ failure, which can be life-threatening. To try and reduce this risk, a team of researchers tinkered with a portion of a receptor found on the surface of the T cells. They then used the modified cells to treat a small number of patients with B-cell lymphoma, whose cancer had either not responded to treatment or had recurred. The 25 people involved received a range of T-cell dosages, and none of them reported serious side effects. 6 of 11 patients who received a higher dose of the modified T cells even went into complete remission. You can find that research in Nature Medicine.

[Jingle]

Host: Shamini Bundell

So, on today’s podcast, I’ve got a story from last week’s Nature, but this was such fascinating research that I decided to cover it anyway. Plus, it gave me the perfect opportunity to set you a little challenge, if that’s okay, Ben?

Host: Benjamin Thompson

Absolutely. Shamini, though, is this one of these challenges where you humiliate me in front of the listeners? I do love those.

Host: Shamini Bundell

Ben, I would never do that. This is a really easy listening challenge. Okay, I am going to play you a sentence and you just have to tell me what it says.

Host: Benjamin Thompson

Okay, let’s do it then.

[Recording]

Host: Benjamin Thompson

Once more?

[Recording]

Host: Benjamin Thompson

So, I think the first bit was ‘rise and shine’, but after that things did get a little fuzzy, to be honest with you.

Host: Shamini Bundell

Okay, well, I’ll play that once more for you.

[Recording]

Host: Benjamin Thompson

‘Rise and shine’… something about Ollie?

Host: Shamini Bundell

Laughs. Well, I’ll play you the original sentence that that was sort of taken from.

Bright sunshine shimmers on the ocean.

Host: Benjamin Thompson

Oh yeah, there it is.

Host: Shamini Bundell

Okay, maybe it wasn’t quite as easy a listening task as I made out. Will you be up for one more?

Host: Shamini Bundell

Always, let’s do it, come on.

[Recording]

Host: Benjamin Thompson

Okay, ‘the print that I would see in’… let’s have one more of that, Shamini. Once more, please.

[Recording]

Host: Benjamin Thompson

‘The print that you are seeing is something, something hooves’.

Host: Shamini Bundell

That one wasn’t bad actually!

The proof that you are seeking is not available in books.

Host: Benjamin Thompson

You’re very generous – I wasn’t even close, to be honest with you, but okay, nice one!

Host: Shamini Bundell

Laughs. Well, they are quite tricky to understand, to be fair, those sentences, but they were produced in a fascinating way. So, what you’re hearing there has been synthesised by a computer, based on readings of a person’s brain signals as that person speaks the sentence out loud, which is what you heard afterwards. The researchers have used neural networks and machine learning to train a computer to turn electrical activity in a particular brain region into artificial speech. The brain region is the area which controls speech muscles, and the ultimate goal is that people who can’t move these muscles, perhaps due to paralysis, could communicate via a brain-computer interface. I called up one of the authors of the paper, Gopala Anumanchipalli, to have a chat about the work, and it turns out a simple chat is quite a complicated affair.

Interviewee: Gopala Anumanchipalli

In normal conversational speech, we talk at about 150 words per minute, upwards of that, and when we talk, we are not just moving one joint or one muscle. We are moving this whole apparatus of your lips and your vocal tract, which involves the tongue, the jaw, and upwards of 100 muscles and joints.

Host: Shamini Bundell

And for some people, using those muscles isn’t possible or isn’t easy. What kind of things currently exist to help people communicate if they can’t use their mouth and tongue easily?

Interviewee: Gopala Anumanchipalli

The current assistive devices include devices that sense your eye gaze, twitch of your muscle, to move a cursor around the screen to select letters and string words together, and that’s then sent off to a speech synthesiser much like Professor Hawking used to do. And given how slow it is, we realise how extremely hard it gets to communicate.

Host: Shamini Bundell

Because if you’re having to spell out each individual letter, the number of words you can say in a minute is much lower than the sort of rapid-fire conversation that most people use.

Interviewee: Gopala Anumanchipalli

Right now, you and I are speaking at over 150 words and we are not thinking about how complex engineering the vocal tract has.

Host: Shamini Bundell

So, your research, in a way, aims to replicate some of that complex engineering that allows the brain to control the vocal tract which produces speech. The ultimate aim would be an assistive device which is way more fluent and easy to use than moving a cursor from one letter to the next. Instead, you could take an output from a person’s brain and take it directly into speech, so kind of like reading their thoughts.

Interviewee: Gopala Anumanchipalli

So, this is not a technology which would read thoughts. Firstly, we do not know how to do that, we did not want to do that, but that may not be necessary for the target that we have right now which is like helping these people with neurological disorders communicate because the parts of the brain that they are still intact with include the intent to move and that’s what we are tapping in to.

Host: Shamini Bundell

So, rather than trying to read people’s inner thoughts, the goal would be that people who can’t move still have the brain areas that control movement. So, if you can tap into almost what they’re trying to say, you could then enable them to communicate that way.

Interviewee: Gopala Anumanchipalli

That is the goal. So, we do not know at this point what is the limit of what’s plausible from reading out the brain, so we needed a proof of principle to show that it’s possible to decode speech directly from the brain activity.

Host: Shamini Bundell

So, this experiment is a sort of very fascinating step towards that ultimate goal. It’s not about people who can’t speak, but it’s about training a computer to understand what people are trying to say. So, how did you go about doing that?

Interviewee: Gopala Anumanchipalli

We work in the department of neurosurgery. We have the epilepsy monitoring unit here where we treat people with seizures that do not go down with meds, so they have to result to neurosurgery where the skull is taken off and this grid of electrodes essentially are put on the surface of the cortex. And part of our experiments include like can you speak these sentences naturally, and they would produce speech and we collect both the audio and the neural data in synchrony so we can correlate one against the other and attempt to transform one representation to the other.

Host: Shamini Bundell

And previous work has looked at signals in the brain from electrodes, from patients like this, and tried to say okay, when they’re saying something, they get these signals and this output and link them directly. But you took a slightly different approach in this study because you were particularly interested in how the brain signals were controlling the muscle and vocal tract movements, right?

Interviewee: Gopala Anumanchipalli

That’s right. So, one way to go about this would be like you have your brain data, you have your audio data and then you build neural networks to convert one form of representation to the other. But what we’ve done is we had an earlier study where we have shown that in this part of the brain which is responsible for controlling the vocal tract, it’s all about movement, rather than the sounds it produces, and hence we figured if that’s what the brain is doing, we should do the same in going from brain to speech, we should go from to brain to what the closest representation in the brain is, which is the articulatory movements, which include movements of the lips, the tongue, the jaw and the larynx and so forth, and then these are converted acoustic signals that we can hear.

Host: Shamini Bundell

And the neural networks had sort of quite a lot of sentences that these patients were saying so that they could be trained on enough data that they could work out the kind of pattern match between, okay, if the brain is making these patterns then the mouth is making these patterns and this is the sound, and I played some of the outputs of the speech synthesiser to my colleague and I’d say they were quite tricky to match but you could definitely recognise the basic sounds and shapes being made that were matching there.

Interviewee: Gopala Anumanchipalli

I think that’s the level intelligibility that we’re at right now, but what we also found is context is extremely important. In normal conversations, we are always talking in context and we are hearing in context and that’s what we think makes the current performance as is helpful.

Host: Shamini Bundell

And so, your experiment was done with five patients who are undergoing surgery for epilepsy and therefore had these electrodes in their brains and they were all people who were able to talk, so you trained the machine as they talked out loud. How could this translate into a situation where someone couldn’t talk out loud.

Interviewee: Gopala Anumanchipalli

One of the things we wanted to try is have these people silently mouth these sentences which would mean moving all the right muscles and the jaw and so forth, so there is no audio output that we are recording but just the brain function that underlies these movements. And what we found is that although it’s worse than audible speech, we heard something that was actually speech-like. The scientific unknowns are how well this translates to someone who has lost speech function, and the only way to find out would be to try it out. Yeah, we are looking forward to seeing where this research goes and we are actively excited about it.

Host: Shamini Bundell

That was Golapa Anumanchipalli of the University of California San Francisco. You can find the paper, a News & Views commentary and a news article on it over at nature.com.

Interviewer: Benjamin Thompson

Finally on this week’s show, it’s time for the News Chat and I’m joined here in the studio by Matthew Warren, one of the reporters here at Nature. Matthew, thanks for dropping by.

Interviewee: Matthew Warren

Hi, nice to be here.

Interviewer: Benjamin Thompson

Our first story then, Matthew, we’re going to be talking about Denisovans, this mysterious ancient hominin group. What’s been going on?

Interviewee: Matthew Warren

Yeah, so since they were discovered a decade ago, researchers have been looking for more remains of these Denisovans. So far, they’ve only found them in this one cave in Siberia and finally they’ve found this new jawbone from Tibet, which seems to be a Denisovan.

Interviewer: Benjamin Thompson

Well, what can you tell me about the Denisovans then sort of in general?

Interviewee: Matthew Warren

Well, we don’t know very much about them. They’re a fairly mysterious group of kind of extinct hominins, but we know that modern humans in places like Asia and Oceania have some Denisovan DNA in their genome, and so for a while now, researchers have expected that these Denisovans lived kind of far beyond this one cave in Siberia where they’ve been found.

Interviewer: Benjamin Thompson

Right, and this find is maybe evidence of that then, so this is Tibet which is thousands of miles away from the original cave.

Interviewee: Matthew Warren

Yeah, that’s right and people have thought maybe China would be the place to find these other Denisovan fossils and it turns out they were right. But interestingly, it was a fairly unusual location on the Tibetan Plateau, over 3,000 metres above sea level. So, that’s really high for a hominin group to be living in.

Interviewer: Benjamin Thompson

How do we know though that this is a Denisovan and not maybe a Neanderthal or something like that?

Interviewee: Matthew Warren

Well, that’s one of the most interesting things about this study. So, previously, people have identified Denisovan remains by looking at ancient DNA that has remained in the fossils, but this fossil didn’t have any DNA. Researchers couldn’t extract any DNA from it so instead they looked at the ancient proteins that were harboured within the fossil, and this was the first time that people have identified an extinct hominin solely on the basis of the proteins that remain in there.

Interviewer: Benjamin Thompson

On a previous News Chat, Matthew, we talked about the potential of using proteins to try and classify sort of ancient hominin species, but it sounds like it’s been done then in this case. What sort of proteins were the researchers looking at?

Interviewee: Matthew Warren

In this case, the researchers were specifically looking at collagen proteins from the teeth of the specimen. So, collagen proteins make up the structure of the teeth. But the study kind of suggests that maybe ancient protein analysis could be used for other similar specimens in the future. In other cases where researchers are unable to extract DNA, maybe they’ll still be some proteins surviving and maybe they could use that to place a specimen into an evolutionary tree like this. And that could be particularly useful for older specimens where there might still be proteins even though the DNA has degraded or specimens that have been uncovered in warmer climates like southeast Asia, for example, because in those climates DNA degrades relatively quickly but again, proteins might stand more of a chance of surviving.

Interviewer: Benjamin Thompson

So, the researchers have been able to identify this jawbone then as being from a Denisovan – what’s it telling researchers in general about the movement of hominins in ancient times?

Interviewee: Matthew Warren

Well, for a start it kind of proves researchers correct that these Denisovans did venture far beyond just Siberia, and one of the things it suggests is that Denisovans lived at surprisingly high altitude and one of the things about living at such a high altitude is that you have lower oxygen levels. And actually, we already knew that modern-day people living in Tibet have received a version of a gene that helps them live at those high altitudes and we knew that that came from Denisovans but that was really hard to reconcile with the fact that Denisova Cave in Russia – where all of the other Denisovans have been found so far – is only 700 metres above sea level, so not at a very high altitude. So, this new study suggests that maybe Denisovans lived in the Tibetan Plateau for a really long time and developed this gene that helped them live at high altitudes and then later on when Homo sapiens came and arrived at that plateau, the two different groups of humans interbred and that’s why modern-day Tibetans have that gene.

Interviewer: Benjamin Thompson

Well, so we know a bit more about these sort of mysterious Denisovans now, Matthew – where do we go next with this one?

Interviewee: Matthew Warren

Well, this is the most sort of morphologically informative specimen they’ve found so far, so the specimen that can tell us most about what Denisovans looked like. Previously, the other remains that have been found at Denisova Cave have all been teeth or shards of bone. And so, now that we’ve got this half jawbone, scientists can begin comparing it against other fossils which have already been uncovered. So, there are some in China, for example, that look similar in many ways to this fossil and so it could be that those are Denisovans and so in the future I think scientists would like to go and do DNA analysis or protein analysis on those as well and figure out if they too could be Denisovans.

Interviewer: Benjamin Thompson

Well, let’s move on to our second story today then Matthew and, well, it couldn’t be more different really. This one is looking at how the words used by male and female scientists in grant proposals is maybe affecting their outcomes.

Interviewee: Matthew Warren

Yeah, that’s right. So, foundations that offer grants to scientists will often blind their review process so that the reviewers don’t know if an application was submitted by a man or a woman. But a new study has come out that suggests that even when that process is blinded, there is still a bias which seems to come down to the language that is used in applications from men versus women.

Interviewer: Benjamin Thompson

Right, well how was this done then?

Interviewee: Matthew Warren

So, researchers looked at almost 7,000 applications to the Bill and Melinda Gates Foundation between 2008 and 2017, and they analysed the wording that scientists used in those grant proposals, and they looked at how often people used broad words to describe the research that they did, so these were words that kind of appeared at the same rate across all of the different proposals, regardless of what topic they were. So, some of the examples they give are things like ‘control’ or ‘detection’. And they also looked at the number of words that were more kind of topic-specific, so these were ones which were relevant just to a narrow range of topics, so these were words like ‘brain’ or ‘community’.

Interviewer: Benjamin Thompson

And what did the researchers analysing these words find then?

Interviewee: Matthew Warren

They found that broad words, when they were used in an application, were associated with higher scores from the reviewers, and narrow words were associated with lower scores. But there was also a gender difference here, so men tended to use those broader words, whereas women tended to use more of those narrow words. This kind of meant that women were receiving lower scores overall on these grant proposals than men were. But the interesting thing is that scientists who used more of those broad words in their research proposals didn’t necessarily have better outcomes later on. It didn’t lead to more publications or a higher number of future grants, for example.

Interviewer: Benjamin Thompson

And this seems to then have implications for the grant reviewing process. You mentioned it was blinded – what are the Bill and Melinda Gates Foundation saying about this result?

Interviewee: Matthew Warren

Well, they say that they’re committed to ensuring gender equality and that they do use blinded reviews in an attempt to eliminate bias. But they also said that they’ll be carefully reviewing the results of the study as part of their ongoing commitment to learning and evolving as an organisation. And the researchers involved in this work also have several ideas about how you could eliminate this bias, so they suggested that maybe grant reviewers could be trained to limit their sensitivity to these different kinds of communication styles. They also suggested that the makeup of the review panel could be important, and so they actually showed that female reviewers’ scores don’t favour the proposals from male applicants in the same way that male reviewers’ scores do, and so they suggest that maybe you could increase the number of female reviewers on those review panels.

Interviewer: Benjamin Thompson

Well, thank you for joining me, Matthew. Listeners, as always, head over to nature.com/news for more on these stories.

Host: Shamini Bundell

That’s it for this week’s show If you’d like to get in touch with us, you can tweet at us – we’re @NaturePodcast – or send us an email – podcast@nature.com. I’m Shamini Bundell.

Host: Benjamin Thompson

And I’m Benjamin Thompson. See you next time.