Models can predict astonishing things about the human liver

Uncovering interactions and external influences

Dr Matthias König, theoretical biophysicist and LiSyM Junior Group Leader at the Institute for Biology, Humboldt University, in Berlin, develops computer models of the human liver. His simulations help quantify individual variations in liver function and the external factors influencing it. With these models, therapeutic drug doses can be optimized; measurements become more meaningful. “These models have much potential,” König says. One such model has shown that smoking alters the results of the LiMAx liver function test. A new clinical trial has been initiated specifically to substantiate these findings.


LiMAx (Liver Maximum Capacity) – test of the liver function capacity.

It is notably used to support decisions involving liver surgery, and to determine the severity of liver diseases.

FAIR – guidelines for the management of scientific data to ensure they are Findable, Accessible, Interoperable and Reusable.
SBML – Systems Biology Markup Language is a standardized data exchange format for models of biochemical and biological networks.


“The extent to which diet and lifestyle influence liver metabolism is fascinating,” according to König, For example, in liver cells, different enzymes breakdown nicotine and caffeine. Yet, nicotine induces the enzyme that breaks down caffeine. Therefore, in the presence of nicotine, the amount of this enzyme increases. This, in turn, accelerates the breakdown of caffeine, a stimulating alkaloid. Or, as König succinctly puts it: “The more one smokes, the less effect caffeine has.” It may be that smokers subconsciously attempt to counteract the effect. Studies suggest that smokers drink more coffee.

One aim of his models is to recognize and quantify such influences and interactions. “Many foodstuffs and lifestyle habits affect the absorption and elimination of drugs.” When doctors know the extent of these effects, they can respond by adjusting the drug dosages accordingly. Therapies would be more successful; side effects less frequent. Similarly, individual variations in liver function could be determined from test results. The subsequent diagnosis would become more meaningful. In addition, König’s model allows as precise a representation as possible of a host of liver functions. Already it can accurately predict how well the organ can metabolize specific substances and which factors play a role in their metabolism.

The multi-scale model of the liver guarantees reproducible results with different simulators. It is a valuable tool for validating certain dynamic liver tests and their results.

Before developing a model, a thorough literature research is necessary. As computers cannot serve this task, König and his team of three need to sieve through findings on the enzymes, inhibitors, co-factors and so on, involved in the processes more or less by hand. “It is very labor-intensive,” he confirms. “We collect everything that needs to be included in the equations.” There are also a few ground rules to consider, such as saturation effects for enzymatic reactions. Using various strategies, the resulting equations are then combined to create a basic model. “We check their performance with training data, which have already been entered during model development.” König explains.

Afterwards, newly acquired test data are fed into the simulations. The simulations must be able to reproduce the corresponding processes and results as accurately as possible. “This reveals the model’s predictive capacity, that is, its predictive power,” the biophysicist adds. The model is expanded and refined further through additional equations, data and trial runs. The opportunities expand, but so too does the complexity: “The number of equations increases dramatically!”


Models can enhance liver test results

Models like this can improve the meaning of test results: an assessment of test results alone would suggest that larger livers function better than smaller ones. However, in the latter, only fewer enzymes are at work. “A smaller liver works slower, but relative to its size, by no means any worse,” König explains. In co-operation with doctors from the Charité Hospital, Berlin, König’s model was able to identify a confounding factor in the LiMAx test: “It predicted that smoking influences the results.” First experimental results appear to confirm this finding. To substantiate this, the Charité has recently launched a new clinical study. König is pleased. “That is also acknowledgement for our model. I hope the study is successful.”

Even small details can determine what diagnostic values really mean, or how well and for how long drugs will be effective. König gives an example: “Immediately after eating, the circulation in the liver increases, leading to a change in its metabolic performance.” Liver functions differ from individual to individual due to their different genetic make-up. Diet and other life-style factors accentuate these differences. Therefore, König plans to expand his model. He aims to simulate combinations of two and three substances as well as larger cocktail mixes to discover new interactions and more complex connections.


Absence of data standards creates a lot of work

The LiSyM network provides a wealth of data on single substances and pharmaceuticals and other substance combinations. König aims to collate much of these data. “Working with published data is an enormous, underestimated challenge.” Many data records prove to be incomplete, incorrect or cannot be used. “Standardization would save us an awful lot of work,” says König. For this reason, among other things, he promotes the management of data according to the FAIR guidelines and the SBML standard for models. These require several routine tasks: finding data sets, adapting and weighing them according to their relevance, feeding the data into the model and then connecting computers into specialized clusters to create sufficient processing power.

The expert believes it worth the effort: “Our multi-scale models have the huge advantage of representing everything – from single enzymes, through to tissue structure and blood flow and ultimately the whole body.” After completing his PhD. in theoretical biophysics at the Charité, König returned to the Humboldt University, in Berlin, where he had previously studied biophysics. Today he supervises his LiSyM Junior Group. He hopes that, one day, his models will be used routinely in prevention and diagnosis, and to monitor the effectiveness of therapies. Matthias König stresses: “Computer models are very useful tools. I could even imagine founding a start-up business with them.”

Back to the junior groups