Thomas Schlichthärle

Thomas Schlicht­här­le sees biology as a system to be op­ti­mi­zed. The system com­pon­ents that he’s par­ti­cu­lar­ly in­te­rested in are all the pro­te­ins that perform vital tasks in human cells. When they mal­func­tion, it can lead to di­sea­ses. 

“If we un­der­stand the me­cha­nisms of these protein ma­chi­nes and why they are de­fec­tive, we can repair them or even build better pro­te­ins than the ones pro­du­ced in nature,” says Schlicht­här­le. Instead of using Petri dishes, pi­pet­tes, and mi­cro­scopes, this bio­che­mist employs the latest ar­ti­fi­ci­al in­tel­li­gence (AI) models. His spe­cia­list field, protein design, sits at the in­ter­face of bio­che­mi­stry, struc­tu­ral biology, me­di­ci­ne, and com­pu­ter science. It is the art of imi­ta­ting nature to create pro­te­ins that perform a de­si­ra­ble new func­tion, so as to steer cell pro­ces­ses towards healthy out­co­mes.

“AI has ac­ce­le­ra­ted the pos­si­bi­li­ties for protein design dra­ma­ti­cal­ly. We can now simply try so­me­thing out, design a protein, and predict its three-di­men­sio­nal struc­tu­re – and the­re­fo­re its func­tion – on the com­pu­ter,” says Schlicht­här­le, who took up the Chair for AI-guided Protein Design at the Tech­ni­cal Uni­ver­si­ty of Munich (TUM) in June 2025. “Then we can see whether the protein will be well folded or not, and can make further im­pro­ve­ments to it with the help of AI models, before pro­du­cing it in the lab and testing it to see whether it func­tions as in­ten­ded.”

For in­stan­ce, he and other protein de­si­gners have suc­cee­ded in re­pro­gramming nerve growth factor (NGF) in such a way that it con­ti­nues to fulfill the task that medics want it to perform – namely re­pai­ring damaged nerves – but without causing pain. To do this, the re­se­ar­chers de­ve­lo­ped a new protein that no longer binds to a pain re­cep­tor but other­wi­se ex­hi­bits the same ac­tivi­ty as the ori­gi­nal NGF. “It can be used any­whe­re where pe­ri­pheral nerves have been damaged and need to be re­ge­ne­ra­ted,” says Schlicht­här­le. And the pa­ti­ents don’t feel any pain. 

30,000 po­ten­ti­al protein struc­tures for one perfect can­di­da­te

So how exactly do you go about making an op­ti­mi­zed protein? Schlicht­här­le’s team follows a method that in­vol­ves three dif­fe­rent AI models. The first one, a dif­fu­si­on model, ge­ne­ra­tes up to 30,000 protein struc­tures on the basis of un­st­ruc­tu­red data and certain in­st­ruc­tions – such as the in­st­ruc­tion not to bind to a par­ti­cu­lar pain re­cep­tor. A second AI model finds the un­der­ly­ing amino acid se­quence which ensures that the protein folds into its three-di­men­sio­nal struc­tu­re. Finally, a power­ful struc­tu­re-pre­dic­tion model, Al­pha­Fold, tests the ef­fec­tiveness of the de­si­gned pro­te­ins. 

“Using in-house pro­grams, we select the best pro­te­ins and feed them back into the first model,” Schlicht­här­le ex­p­lains. After two or three passes, the re­se­ar­chers are left with 96 pro­mi­sing protein can­di­da­tes that can then be pro­du­ced in the lab with the aid of bac­te­ria and tested to see how they func­tion.

From am­bu­lan­ce driver to protein de­si­gner

Schlicht­här­le came to protein design through me­di­ci­ne. He decided back when he was driving an am­bu­lan­ce during his gap year that he wanted to help people – but as a re­se­ar­cher, not as a doctor. He studied mole­cu­lar me­di­ci­ne in Tü­bin­gen, fol­lo­wed by bio­en­gi­nee­ring in Dresden. He wrote his master’s dis­ser­ta­ti­on on DNA na­no­tech­no­lo­gy at the Wyss In­sti­tu­te for Bio­lo­gi­cal­ly In­spi­red En­gi­nee­ring in Boston. His PhD, which he com­ple­ted at the Max Planck In­sti­tu­te of Bio­che­mi­stry in Mar­tins­ried, focused on la­be­ling pro­te­ins with mole­cu­lar “tags” so that they can be iden­ti­fied under high-re­so­lu­ti­on mi­cro­scopes. 

He pro­gram­med ma­the­ma­ti­cal si­mu­la­ti­ons to close the gaps. Then, in 2017, he heard about the pos­si­bi­li­ties offered by protein design, in a lecture given by US bio­che­mist David Baker. “I im­me­dia­te­ly knew that we could build much better tags using his methods,” he says. 

Schlicht­här­le spent nearly five years as a postdoc working in Baker’s lab at the Uni­ver­si­ty of Wa­shing­ton. He ab­sor­bed the protein design methods like a sponge. In 2024, during his last year in Seattle, he woke up at four o’clock one October morning, looked at his phone, and saw the news: The Nobel Prize in Che­mi­stry had been awarded to David Baker for his achie­ve­ments in com­pu­ta­tio­nal protein design. The other half of the prize went to Demis Hass­a­bis and John Jumper at Google De­ep­Mind for the de­ve­lop­ment of Al­pha­Fold2, an AI model that can predict the 3D struc­tu­re of pro­te­ins from their amino acid se­quen­ces. 

“After the press con­fe­rence, David came into the lab and we ce­le­bra­ted tog­e­ther,” says Schlicht­här­le. It was clear to all of them that the Nobel Prize was a be­gin­ning, not an end. Protein design had reached the stage where it could be used to tackle all kinds of real-life pro­blems, from de­ve­lo­ping new tre­at­ments to crea­ting new enzymes that can break down the PET in plastic bottles and be de­ploy­ed in other in­dus­tri­al ap­p­li­ca­ti­ons. 

Schlicht­här­le moved from Seattle back to Munich. His time on the West Coast of the United States had pro­vi­ded a mo­ti­va­ti­on boost – partly because of the Nobel Prize, but also because David Baker had made a deep im­pres­si­on on him as an out­stan­ding re­se­ar­cher and leader. “There were around 100 people in his group, but he was always there for us when we had ques­ti­ons, and at luncht­i­me, when he would sit with us in the kitchen to eat his salad,” he recalls. 

A solid basis for a virtual cell

Now, at TUM, Schlicht­här­le plans to use protein design methods to develop highly ef­fi­ci­ent mole­cu­lar tags to label and vi­sua­li­ze 90% of the pro­te­ins present in cells . “We want to make the tags avail­ab­le through a startup and share the as­so­cia­ted protein se­quen­ces openly with the re­se­arch com­mu­ni­ty,” he says. 

Once the foun­da­ti­ons are in place, Schlicht­här­le plans to work with cli­ni­ci­ans to design new pro­te­ins. Pro­te­ins for new cancer im­mu­no­the­ra­pies, for example. Or pro­te­ins that can enable ef­fi­ci­ent pro­duc­tion of spe­cia­li­zed cells from stem cells to test the ef­fec­tiveness of new drugs. He has a future vision of a virtual stem cell that will be sti­mu­la­ted by newly de­si­gned growth factors on a com­pu­ter, instead of in a lab. It will be a kind of digital Petri dish that can predict real cell re­ac­tions. In other words, nothing less than a whole new basis for drug de­ve­lop­ment.