Butterflies of the soul; Neuroscience.

"How neurons really work is slowly being elucidated. That will help both medicine and the search for better artificial intelligence

A NEURON IS a thing of beauty. Ever since Santiago Ramón y Cajal stained them with silver nitrate to make them visible under the microscopes of the 1880s (see drawing above), their ramifications have fired the scientific imagination. Ramón y Cajal called them the butterflies of the soul.

Those ramifications--dendrites by the dozen to collect incoming signals, called action potentials, from other neurons, and a single axon to pass on the summed wisdom of those signals in the form of another action potential, turn neurons into parts of far bigger structures known as neural networks. Engineers now use simulacra of these to create what they are pleased to call artificial intelligence, though it is a pale shade of the real thing.

How neurons actually work their magic is only now being disentangled. One conclusion is that each is, in its own right, as powerful an information processor as a fair-sized artificial neural network. That has implications not only for learning how brains work--and how they go wrong--but also for designing artificial versions that more closely resemble the natural sort.

The first widely adopted neuron model, proposed in its existing form in 1957 by Frank Rosenblatt, an American psychologist (who drew, in turn, on Alan Turing, a British computing pioneer), was the perceptron. This is a mathematical function that receives sets of binary digits (zeros and ones) as inputs. It multiplies these by numerical "weights" and then adds the products together. If the result exceeds a preordained value, the perceptron spits out a "one". If not, it spits out a "zero".

Layer cake

To make artificial neural networks, perceptrons are encoded as software. They are organised, logically speaking, into interconnected layers and the result is trained to solve problems via feedback and feedforward loops between the layers. These loops alter the values of the weights, and thus the behaviour of the network. The more layers, the "deeper" the network. Deep neural networks now underpin everything from Google Translate to Apple's Siri.

All this imitates how action potentials arriving at the synaptic junctions between axons and dendrites, via which neurons communicate, were thought to trigger signals that then combined with each other to trigger (or not) new action potentials in the receiving cell's axon. It is thus tempting to see neurons as physical perceptrons, with the difference from the computer versions that their signals are carried by sodium, potassium and calcium ions crossing cell membranes, rather than by a flow of electrons. And for decades that was just how many neuroscientists did see them.

In the early 2000s, though, Panayiota Poirazi of the Institute of Molecular Biology and Biotechnology in Heraklion, Greece, began looking at the matter differently. She imagined neurons themselves as perceptron networks. In 2003 she argued that a simple two-layer network might be enough to model them. Recent work has upped the ante. In 2021 David Beniaguev of the Hebrew University of Jerusalem concluded that, for human cortical neurons at least, five (and sometimes as many as eight) layers are needed, each with up to 256 perceptrons.

This means lots of computing must be going on inside individual neurons. And it is. Dendrites are now known to generate their own, tiny action potentials, called dendritic spikes. These come in several varieties: calcium spikes (long and slow); sodium spikes (short and fast); and NMDA spikes (triggered by a chemical called N-methyl- D-aspartate). Together, they let dendrites perform 15 of the 16 basic operations of Boolean algebra, a branch of mathematics that is the basis of digital computing. Those operations compare two input values and spit out a third as a result. Some, such as AND, OR, NOT and NOR, are self-explanatory. Others, such as NAND, XNOR and XOR, less so.

XOR, in particular, is notorious. It gives a non-zero output only when its inputs are dissimilar. In 1969 two eminent computer scientists, Marvin Minsky and Seymour Papert, proved that XOR cannot be performed by a single perceptron--one of only two Boolean operation for which that is the case. This result stalled artificial-intelligence research for a decade, the first " AI Winter", as it is retrospectively known.

That was thought true of dendrites, too. But in 2020 work by Albert Gidon of Humboldt University, in Berlin, in which Dr Poirazi was also involved, found a new class of calcium-based spike which permits XOR. That a single dendrite can thus outperform a perceptron suggests an entire layer of complex computation is going on out of sight of conventional models of neurons. That might help explain the remarkable performance of brains and the failure of artificial intelligence to reproduce it.

Axons, too, have been reassessed. The action potentials they carry had once been seen by many as analogous to the all-or-nothingness of a binary digit. Look closely, though, and action potentials vary in both height and width. That matters.

In 2016 a group from the Max Planck Institute for Neuroscience, in Florida (one of the organisation's few campuses outside its German homeland), showed that neurons in the central nervous system actively adjust the breadth of their action potentials. The following year a team from Dartmouth College in New Hampshire discovered that those in the cortex actively adjust their heights as well.

Even the lengths of the intervals between action potentials matter. In May 2021 Salman Qasim of Columbia University reported that neurons in the hippocampus, a part of the brain involved in memory formation, modulate the timing of their firing to encode information about the body's navigation through space. And in August of that year Leila Reddy and Matthew Self of the University of Toulouse, in France, reported that neurons also do this to encode the order of events in memories.

All this has clinical implications. In particular, there is growing evidence that atypical dendrite development in childhood and early adulthood is linked to autism, schizo phrenia and epilepsy. Deteriorating axonal function, meanwhile, is similarly associated with psychosis in multiple sclerosis, schizophrenia and bipolar disorder. These discoveries inform the development of new medicines. For example, ketamine, which triggers long-lasting structural change in dendrites, is receiving attention as a treatment for depression.

The art of forgetting

The sophistication of the neuron and its constituent parts has also caught the attention of computer scientists. In the early 2010s deep neural networks drove such dramatic improvements in the abilities of artificial intelligence that there was genuine concern people would soon have to wrestle with machines cleverer than they were. Then, suddenly, progress stalled.

Deep neural networks have hit three obstacles. First, computer scientists found that once a network has learnt a task, it struggles to transfer those skills to a new one, however similar, without extensive retraining. Second, when such a network is retrained, it tends to forget how to perform the original task--an effect called catastrophic forgetting. Third, to train a large network requires immense volumes of data, access to supercomputers, and the megawatts of electricity needed to run those supercomputers for days (or even weeks) at a time.

The brain struggles with none of this. It effortlessly transfers knowledge between domains, has no trouble integrating old and new skills, and is remarkably efficient--running on watts, not megawatts. The sophistication of neurons may make the difference. In studies published last year and this, a team from Numenta, a Californian research company, designed artificial neurons, with dendrite-like subcomponents, that are immune to catastrophic forgetting. A network of these trained on 100 tasks in sequence retained the ability to perform all with reasonable accuracy. The same network also outperform networks of perceptrons at learning many tasks simultaneously.

Several studies show that sophisticated artificial neurons can approximate complicated functions-- XOR, for example--with greater accuracy and less energy than perceptrons do. Connected into networks, such devices learn faster and at a lower computing cost than perceptrons. The question of how brains apply knowledge from one domain to others remains a mystery, but it would not be a surprise if the complexity of neurons explains that, too.

The lesson, then, is familiar: nature got there first. Necessity may be the mother of invention, but natural selection is the mother of inventors. In both neuroscience and artificial intelligence the next decade promises to be wild. Over a century after he described them, Ramón y Cajal's butterflies are taking flight." [1]

 

·  ·  · 1.  "Butterflies of the soul; Neuroscience." The Economist, 2 July 2022, p. 72(US).

 

EU founders also will get control: EU deep technology strategy - 45 billion Eur funding, more favorable IPO rules

"On Tuesday, the European Commission (EC) should present a strategy for the promotion of deep technologies, the main axis of which is more favorable IPO rules for startups operating in this area.

 

To promote Deeptech in the European Union (EU), they want to attract 45 billion EUR of funds from the private sector, and to adjust the IPO rules so that after listing the company on the stock exchange, it would be easier for the founders to retain a controlling stake.

 

The EU's proposed act "should allow founders and their families to retain control of the company after listing, raising more funds and taking advantage of the company's listing," according to a new digital innovation strategy plan seen by Bloomberg News.

 

The EU lags far behind the US and China in terms of the number of deep technology companies - according to the EC, this is due to the fragmented venture capital markets in Europe and the fact that traditional banks play too large a role in financing start-ups, rather than alternative sources of funds.

 

The plan includes several key points highlighted by Bloomberg.

 

For example, in addition to simplifying IPO rules, it is believed that the EC will propose to EU countries to unify the regulation of dual class share structures. Now individual stock exchanges determine regulation individually.

 

It will also be proposed to allow state aid for the development of test infrastructure, to promote regulatory "sandboxes" where startups could test their ideas.

 

"Deeptech" refers to those technologies based on a high level of innovation and facing large technological and engineering challenges, the solution of which takes a long time. They have a long innovation cycle, a high market entry risk and risks arising from it. Such technologies currently include artificial intelligence, advanced materials, blockchain, etc."

 

 

 

ES steigėjai taip pat turės kontrolę: ES giliųjų technologijų strategija - 45 mlrd. Eur finansavimo, palankesnės IPO taisyklės

 "Europos Komisija (EK) antradienį turėtų pristatyti giliųjų technologijų (angl. Deeptech) skatinimo strategiją, kurios pagrindinė ašis – palankesnės IPO taisyklės šioje srityje veikiantiems startuoliams.

 

„Deeptech“ skatinimui Europos Sąjungoje (ES) norima pritraukti 45 mlrd. Eur lėšų iš privataus sektoriaus, o IPO taisykles pakoreguoti taip, kad įmonę listingavus biržoje, įkūrėjams būtų lengviau išsaugoti kontrolinį akcijų paketą

 

ES siūlomas aktas „įkūrėjams ir jų šeimoms turėtų leisti išlaikyti įmonės kontrolę po listingavimo, pritraukiant daugiau lėšų bei pasinaudojant įmonės listingavimo privalumais“, teigiama naujosios skaitmeninių inovacijų strategijos plane, kurį matė „Bloomberg News“.

 

ES pagal giliųjų technologijų bendrovių skaičių gerokai atsilieka nuo JAV ir Kinijos – pasak EK, tai lemia fragmentuotos rizikos kapitalo rinkos Europoje bei tai, jog per didelį vaidmenį, finansuojant startuolius, vaidina tradiciniai bankai, o ne alternatyvūs lėšų šaltiniai.

 

Plane numatyti keli pagrindiniai punktai, kuriuos išskiria „Bloomberg“. 

 

Pavyzdžiui, be IPO taisyklių supaprastinimo, manoma, jog EK pasiūlys ES valstybėms suvienodinti dvigubos klasės akcijų struktūrų reguliavimą. Dabar atskiros akcijų biržos reguliavimą nustato individualiai.

 

Taip pat bus pasiūlyta leisti valstybės pagalbą kuriant testinę infrastruktūrą, skatinti reguliacines „smėlio dėžes“ (angl. Sandbox), kuriose startuoliai galėtų išbandyti savo idėjas.

 

„Deeptech“ vadinamos tos technologijos, grįstos aukšto lygio inovacijomis bei susiduriančios su dideliais technologiniais ir inžineriniais iššūkiais, kurių sprendimui reikia daug laiko. Jos turi ilgą inovacijų ciklą, didelę patekimo į rinką riziką bei iš to kylančias rizikas. Tarp tokių technologijų šiuo metu įvardijamas dirbtinis intelektas, pažangiosios medžiagos, blokų grandinė ir pan.” 

 

How the "Octopus" platform wants to revolutionize research

"In order to avoid errors and false incentives, studies are published on the "Octopus" website in a completely different way than usual. But can this solve the problems in science?

Science is constantly correcting itself. When researchers write a study, other researchers look at it, and then even more so when it's published. What is wrong can be corrected. For several years, a whole strand of science has also been working on creating proposals for a new publication system in the scientific community. Because that also contains certain things - some would say "mistakes", others perhaps only speak of "wrong incentives". Everyone should agree that there is room for improvement. Now the Joint Information Systems Committee (Jisc), a British non-governmental organization that develops and promotes new technologies for research and teaching, has presented a "groundbreaking open research platform": Octopus.

Octopus, so the promise, should enable the fast, free and fair publication of research results that is open to everyone. The platform should change nothing less than the entire incentive structure in science, says Alexandra Freeman of the Winton Center for Risk and Evidence Communication at the University of Cambridge, who leads Octopus. 

 

The platform eliminates the problem of scientific work being judged solely on its results, which can lead researchers to choose impressive-sounding results over sound theory and methodologies. 

 

"It will encourage researchers to specialize in their skills as theorists, methodologists, data collectors, analysts, implementers or critics, rather than trying to do everything at once," Freeman said. 

 

Researchers could thus concentrate more on the quality of their work.

 

In fact, the platform fits into a trend that has been increasingly evident in the open science scene for several years: away from result evaluation towards process evaluation. Good research instead of good results.

 

 In this case, good research means: cleanly derived hypotheses, a sufficient number of test subjects, the appropriate statistical methods, transparent reporting. 

 

It's not as if the conventional publishing industry completely ignores this shift in emphasis. There, too, there are journals that attach great importance to the methodology and, for example, check the structure of a study before the data is collected. There are magazines that publish zero results - i.e. those that turn out differently than previously assumed - or those that make sure that all the data is published so that skeptics can do the math.

Why shouldn't the writing of research snippets also be subject to constraints?

 

But Octopus wants to go one step further and break up the classic article format - introduction, method section, results, discussion - into individual parts. Researchers can therefore publish small individual contributions on the platform in the categories: Problem, peer review, hypothesis, protocol, result, analysis, interpretation and application. 

 

The one big paper that contained all of this no longer exists.

 

 Individual contributions can be rated from zero to five stars by other researchers. 

 

The advantages for researchers are said to be that they can publish their articles more quickly and do not have to wait months for a magazine to respond. In addition, they can only publish in areas in which they are experts: one person may be more familiar with methods, the next with transferability to everyday life.

 

But is this groundbreaking? One who can try to assess is Malte Elson, junior professor for the psychology of human-machine interaction at the University of Bochum, he does a lot of research on the research itself. "The page describes a rather ambitious goal, namely an alternative to the classic paper," says Elson. "What's not entirely clear to me: What problem does it solve?"

With large research projects it can certainly make sense not to wait until the very end with publications. "But that is already possible, for example in dedicated technical journals for methods," says Elson. "The benefit of an early publication of a hypothesis - which, for example, in the social sciences is often not formalized but consists of only a few sentences - detached from any theoretical or empirical work, is not immediately apparent to me." In addition, according to Elson, it is not entirely clear to him "why the writing of research snippets is not subject to exactly the same constraints as the writing of a classic essay".

It is not yet certain that the rest of the professional world will also consider the project as groundbreaking as those responsible for Octopus do themselves, and that Octopus will revolutionize the publication system. But maybe it doesn't have to be. A few advantages result from this as well.

He likes the basic idea of making research processes transparent, says Elson. "In this way, the public can be informed very clearly about how research works, from the initial idea through to application." The project is also interesting for the further training of the researchers themselves and especially their offspring. "I could well imagine using it in teaching future scientists," says Elson."

Kaip „Octopus“ platforma nori pakeisti mokslinius tyrimus

"Siekiant išvengti klaidų ir klaidingų paskatų, „Octopus“ svetainėje tyrimai skelbiami visiškai kitaip, nei įprastai. Bet ar tai gali išspręsti mokslo problemas?

Mokslas nuolat taiso save. Kai mokslininkai aprašo tyrimą, kiti tyrinėtojai žiūri į jį, o tada dar labiau, kai jis paskelbiamas. Kas negerai, galima pataisyti. Jau kelerius metus visa mokslo kryptis taip pat kuria pasiūlymus dėl naujos publikavimo sistemos mokslo bendruomenėje. Nes tame irgi yra tam tikrų dalykų – vieni sakytų „klaidos“, kiti galbūt tik apie „neteisingas paskatas“. Visi turėtų sutikti, kad yra kur tobulėti. Dabar Jungtinis informacinių sistemų komitetas (Jisc), Didžiosios Britanijos nevyriausybinė organizacija, kurianti ir propaguojanti naujas mokslinių tyrimų ir mokymo technologijas, pristatė „novatorišką atvirą mokslinių tyrimų platformą“ – „Octopus“.

„Octopus“, taigi pažadas, turėtų sudaryti sąlygas greitai, nemokamai ir sąžiningai skelbti tyrimų rezultatus, atvirus visiems. Platforma turėtų pakeisti ne mažiau, kaip visą paskatų struktūrą moksle, sako Alexandra Freeman iš Kembridžo universiteto Wintono rizikos ir įrodymų komunikacijos centro, vadovaujanti „Octopus“. 

 

Platforma pašalina problemą, kai mokslinis darbas vertinamas tik pagal jo rezultatus, todėl mokslininkai siekia pasirinkti įspūdingai skambančius rezultatus, o ne pagrįstą teoriją ir metodikas. 

 

„Tai paskatins tyrėjus specializuotis jų, kaip teoretikų, metodologų, duomenų rinkėjų, analitikų, įgyvendintojų ar kritikų, įgūdžiuose, o ne bandyti viską daryti iš karto“, – sakė F. Freemanas. 

 

Taigi mokslininkai galėtų daugiau dėmesio skirti jų darbo kokybei.

 

Tiesą sakant, platforma atitinka tendenciją, kuri jau keletą metų vis labiau ryškėja atvirojo mokslo scenoje: nuo rezultatų vertinimo pereiti prie proceso vertinimo. Geri tyrimai vietoj gerų rezultatų. Šiuo atveju geras tyrimas reiškia: švariai išvestas hipotezes, pakankamą tiriamųjų skaičių, tinkamus statistinius metodus, skaidrią ataskaitų teikimą. Atrodo, kad įprastinė leidybos pramonė visiškai nepaiso šio akcento pasikeitimo. Ten taip pat yra žurnalų, kuriuose labai svarbi metodika ir, pavyzdžiui, prieš renkant duomenis patikrinama tyrimo struktūra. Yra žurnalų, kurie skelbia nulinius rezultatus, t. y. tuos, kurie pasirodo kitaip nei manyta anksčiau, arba tų, kurie užtikrina, kad visi duomenys būtų paskelbti, kad skeptikai galėtų suskaičiuoti.

Kodėl tyrimo fragmentų rašymui taip pat neturėtų būti taikomi apribojimai?

Tačiau „Octopus“ nori žengti dar vieną žingsnį ir išskaidyti klasikinį straipsnio formatą – įvadą, metodų skyrių, rezultatus, diskusiją – į atskiras dalis. Todėl mokslininkai platformoje gali skelbti nedidelius individualius indėlius kategorijose: problema, tarpusavio peržiūra, hipotezė, protokolas, rezultatas, analizė, aiškinimas ir taikymas. Vieno didelio popieriaus, kuriame visa tai buvo, nebėra. Kiti mokslininkai individualius indėlius gali įvertinti nuo nulio iki penkių žvaigždučių. Teigiama, kad mokslininkų pranašumai yra tai, kad jie gali greičiau paskelbti savo straipsnius ir nereikia laukti mėnesių, kol žurnalas atsakys. Be to, jie gali skelbti tik tose srityse, kuriose yra ekspertai: vienas žmogus gali būti labiau susipažinęs su metodais, kitas – su perkėlimu į kasdienį gyvenimą.

Bet ar tai novatoriška? Tai gali pabandyti įvertinti Malte Elson, jaunesnysis Bochumo universiteto žmogaus ir mašinos sąveikos psichologijos profesorius, jis daug tyrinėja patį tyrimą. „Puslapyje aprašomas gana ambicingas tikslas, būtent alternatyva. į klasikinį mokslo straipsnį“, – sako Elsonas. „Kas man ne visai aišku: kokią problemą tai išsprendžia?

Vykdant didelius mokslinių tyrimų projektus tikrai gali būti prasminga nelaukti iki pat publikacijų pabaigos. „Tačiau tai jau įmanoma, pavyzdžiui, metodams skirtuose techniniuose žurnaluose“, – sako Elsonas. „Ankstyvos hipotezės paskelbimo nauda, ​​kuri, pavyzdžiui, socialiniuose moksluose dažnai nėra formalizuota, o susideda tik iš kelių sakinių, atskirtos nuo bet kokio teorinio ar empirinio darbo, man nėra iš karto akivaizdi. Be to, anot Elsono, jam nėra iki galo aišku, „kodėl tyrimo fragmentų rašymui netaikomi lygiai tokie patys suvaržymai, kaip klasikinės esė rašymui“.

Tai yra, dar neaišku, ar likęs profesionalų pasaulis taip pat laikys projektą, kaip novatorišką, kaip tai daro patys už „Octopus“ atsakingi asmenys, ir ar „Octopus“ pakeis publikavimo sistemą. Bet gal taip ir neturi būti. Tai taip pat turi keletą privalumų.

Elsonas sako, kad jam patinka pagrindinė idėja, kad mokslinių tyrimų procesai būtų skaidrūs. „Tokiu būdu visuomenė gali būti labai aiškiai informuojama apie tai, kaip veikia tyrimai, nuo pradinės idėjos iki taikymo“. Projektas taip pat įdomus tolesniam pačių mokslininkų ir ypač jų atžalų mokymui. „Gerai galėčiau įsivaizduoti, kaip tai panaudosiu, mokydamas būsimus mokslininkus“, – sako Elsonas."

Curevac iškelia koronaviruso vakcinos gamintojui Biontech ieškinį

"Tiubingeno (Vokietija) bendrovė kaltina konkurentą keturių patentų pažeidimu. Curevac iki šiol nesugebėjo sukurti savo vakcinos. 

 

Tiubingeno bendrovė „Curevac“ padavė ieškinį koronaviruso vakcinos gamintojui „Biontech“ ir dviem dukterinėms įmonėms. „Curevac“ mano, kad jos patentus pažeidžia konkurento milijardus kartų parduota koronaviruso vakcina. 

 

Anot Curevaco, kalbama apie adekvačią kompensaciją už intelektinės nuosavybės teisių pažeidimą. Ieškinys buvo pateiktas Diuseldorfo apygardos teismui. Tai apie keturis patentus. „Biontech“ iš pradžių nenorėjo komentuoti, tačiau paskelbė pareiškimą.

Proceso metu turi būti išaiškinta, kokiu mastu atitinkami patentai buvo panaudoti, kuriant Biontech vakciną, sakė Curevac vadovas Franz-Werner Haas. Kokią sumą numato „Curevac“, jis nesakė. Tai turi būti išaiškinta proceso metu. Po daugybės diskusijų šalys šiuo klausimu nesutaria. Pasak Curevac, ji nesiekia įsakymo ir neketina imtis jokių teisinių veiksmų, kurie galėtų kliudyti Biontech vakcinos Comirnaty gamybai, pardavimui ar platinimui.

Po daug žadančios pradžios 2021 m. viduryje Tiubingene įsikūrusiai bendrovei nepavyko pateikti į rinką koronanos vakcinos ir dabar ji nori tapti naujos antrosios kartos vakcinos kūrimo pionieriumi kartu su savo britų partneriu Glaxosmithkline. Klinikinis jos naujojo kandidato tyrimas buvo pradėtas kovo pabaigoje. 

 

Curevac teigia, kad bendrovės intelektinė nuosavybė apsaugo keletą išradimų, kurie laikomi esminiais Biontech kuriant ir kuriant koronaviruso vakcinas. Pavyzdžiui, tai susiję su mRNR molekulių gamyba.

 

Curevac, kaip ankstyvas mRNR technologijos pradininkas, palaikė greitą vakcinų kūrimą. „Atitinkamai, norint investuoti į pažangias mRNR technologijas ir naujas gyvybę gelbstinčių vaistų klases, intelektinės nuosavybės teisės turi būti pripažįstamos ir gerbiamos už teisingą atlygį.

 

„Curevac“ 2000 m. įkūrė Ingmaras Hoerras. Jis laikomas naujos vakcinacijos technikos atradėju. Kaip ir Biontech ir Moderna vakcinos, ji yra pagrįsta vadinamąja pasiuntinio RNR. 

 

Įmonė įsikūrusi Tiubingene ir joje dirba daugiau, nei 500 žmonių Tiubingene, Frankfurte ir Bostone. Hoerras dėl sveikatos priežasčių pasitraukė iš generalinio direktoriaus pareigų 2020 m. pavasarį."

Curevac Is Suing Corona Vaccine Manufacturer Biontech

"The Tübingen (Germany) company accuses the competitor of infringing four patents. Curevac has so far failed to develop its own corona vaccine.

 

The Tübingen company Curevac has filed a lawsuit against the corona vaccine manufacturer Biontech and two subsidiaries. It sees its patents being infringed by the competitor's billion-fold sold corona vaccine.

 

According to Curevac, it is about adequate compensation for an infringement of intellectual property rights. The lawsuit was brought before the District Court of Düsseldorf. It's about four patents. Biontech initially did not want to comment, but announced a statement.

In the process, it must be clarified to what extent the respective patents went into the development of the Biontech vaccine, said Curevac boss Franz-Werner Haas. He did not say what amount Curevac envisages. This must be clarified in the process. The parties disagree on this issue after numerous discussions. According to Curevac, it is not seeking an injunction and does not intend to take any legal action that could impede the production, sale or distribution of the Biontech vaccine Comirnaty.

After a promising start in mid-2021, the Tübingen-based company failed to bring a corona vaccine onto the market and now wants to be a pioneer in the development of a new second-generation vaccine with its British partner Glaxosmithkline. A clinical study of Curevac's new candidate was started at the end of March.

 

Curevac argues the company's intellectual property protects several inventions considered essential to Biontech's design and development of coronavirus vaccines. These relate to the production of mRNA molecules, for example.

 

Curevac supported the rapid development of vaccines as an early pioneer of mRNA technology. "Accordingly, in order to invest in advancing mRNA technology and new classes of life-saving drugs, intellectual property rights must be recognized and respected in the form of fair remuneration."

 

Curevac was founded in 2000 by Ingmar Hoerr. He is regarded as the discoverer of the new vaccination technique.

 

Like the vaccines from Biontech and Moderna, it is based on so-called messenger RNA. The company is based in Tübingen and employs more than 500 people in Tübingen, Frankfurt and Boston. Hoerr retired as CEO in spring 2020 for health reasons."