To get started, click ‘Enter Tool’ on the Bio Model Analyzer website. You will enter Bio Model Analyzer’s blank canvas screen, with the icons aligned at the top of the screen and ‘Default Model: Version 1’ displayed in the lower right-hand corner.

Drag and drop works the same way for cells, membrane receptors, red variables and grey constants. Click the icon, drag it onto the canvas and release the mouse button to place the variable. Note that the canvas has ‘soft spots’ where the components can be placed and if you are not hovering over a soft spot, you won’t be able to place the component. These soft spots are apparent through highlighting. For instance, if you’re dragging a cell, a grid will highlight when you’ve reached a spot where you can release the cell onto the canvas. Similarly, when dragging receptors onto the cell membrane, you’ll see a circular spot appear highlighted in eight possible locations to attach the receptor around a given cell.
Drawing out the relationships that connect these components is done by firstly clicking on the activating / inhibiting relationship button and seeing it highlighted – this means the arrow is ready to be drawn between variables. Hover over a variable where the relationship starts to see it highlight, click, hold and drag it to another variable until that one is highlighted; release the mouse button and the relationship is drawn.

The zoom operates by using either your mouse wheel or the zoom bar in the top right-hand corner. There is also a panning hand icon; when this mode is on (clicked and highlighted), you can pan around the grid, which you cannot do when the selection tool is on.

Clicking on a variable opens up a dialogue box in the bottom left corner. Here you can specify a name and descriptive range for variables. As usual it is good practice not to include symbols in names of variables (or files). For example, a ‘-‘ symbols in the name of a variable increases the likelihood of getting a failed analysis run as the tool processes subtraction relationships too, specified by ‘-‘. Typical qualitatively descriptive ranges are as follows: 0-1 for ‘off’ and ‘on’; 0-3 for ‘off’, ‘low’, ‘medium’ and ‘high’ concentrations; 1-1 for a ‘low concentration constant’; 3-3 for a ‘high concentration constant’; etc.
If you include variables with different ranges in the same model please have a look at "What happens with variables with different ranges?".

You can leave the default target function, which will entail that your variables get processed as follows: ave(pos) – ave(neg). This function means that the variable gradually changes its value to the difference between the average of the variables having an activating influence on it and the average of the variables having an inhibitive influence on it. This target function is built to optimize the usage of the range of the variable (see below when discussing sum).

Different target functions can be written using basic mathematical operators listed below. The following are typical target functions used in Bio Model Analyzer:

1. A constant’s target function is defined by the constant value that it takes (e.g. 0, 1, 2, etc.). [image:constant target box]

2. Using the default target function for a variable that has only inhibitions is not very useful. In such a case, the average of the positive influences is always 0. Then, if any of the negative influences are on, the variable will want to decrease below 0, if that was possible. Hence, it makes more sense to include the target of the maximal possible value minus the average of inhibitions. [image:only inhibition]

3. Sometimes, the average of positive activations is not appropriate if you desire multiple positive inputs having more of an impact than negative inputs. In such a case, it may make more sense to use the sum of values.
[image:sum activation]
Notice that this produces a very different behaviour. For example, if there is medium level of inhibition (2) and both the activators are in their maximum value of activity (4), then the target of the variable would be maximal (4). However, this does not change at all when the activators decrease to 3 or when the inhibitor increases to 3. The default target function, by contrast, computes the target of 2 and utilizes better the range of 0-4 by setting a target of 4 only when both activators are at their maximal level and the inhibitor is at the minimum level.

4. A common target function corresponds to complexation of two substances. In such a case, the level of the result is the minimum of the levels of the two substances. So the target function is min(var(substance1),var(substance2)).

You can also customize the target function, within the limits of the language defined in Bio Model Analyzer. The user has available the following operators: addition, subtraction, multiplication, division, maximum, minimum, ceiling, floor, and average. For example:

1. max(0,floor((var(a)+var(b))/2 – (var(c)+var(d)+var(e))/3)) : this essentially does the average of the positive and negative influences but adds the floor (lower integer value) to ensure that a target of 2.5 does not lead to instabilisation.

2. min(2-var(a),1)*var(b) : this function is calibrated for variables ranging between 0 and 2. It stipulates that if variable a is either 0 or 1 then the target is the value of b but if variable a is 2 then the target is 0. This models a threshold of inhibition by variable a. Where, as long as the level of a is below the maximum, variable b is copied. But once variable a is 2 it inhibits very strongly.

3. max(0,2-((2-var(a))*max(var(b)-1,0)+1)) : this is one of the most complex target functions we have encountered. It is similar to the previous one except that it includes a more complex condition. Again calibrated for the range 0-2. If variable b is 2 and variable a is not 2 then the target is 0, otherwise the target is the value of variable a. This expresses an inhibition that occurs only when b is at its maximum value and when a is not at its maximum value.

It is important to include in target functions only variables that have a relationship with the variable. If multiple variables of the same name have a relationship with the same variable, Bio Model Analyzer assumes that they affect it in a symmetric way. E.g., in var(a) + var(a) the association between the different a’s and the variables they represent is symmetric. In var(a)-var(a) it is not symmetric. Such an a-symmetric function could lead to problems with analysis. Similarly, if a variable has one relationship with one variable named b, then b cannot appear twice in the target function.

BioModelAnalyzer is built to facilitate the usage of the default target function. What should happen if I have a variable X with range [0,1] activating a variable Y with range [0,2] (and no other influences)? In this case, it makes sense that the maximal value of X should lead to Y having maximal value as well. This means that the ranges of X and Y need to be adjusted when they affect each other. So BioModelAnalyzer does an automatic range conversion when X appears in the target function of Y (or vice versa). If you just use the default target functions this is not something you have to worry about. It essentially means that the entire range of Y will be possible in the example above (and similar cases). However, if you are building your own target functions this may lead to unexpected behaviour.
The exact range conversion that BioModelAnalyzer applies is as follows. If a constant with range [n,n] appears in the target function of a variable with a different range, then no range conversion is applied. If a variable X with range [a,b] appears in the target function of a variable Y with range [c,d], then whenever X appears in the target function of Y it will be modified to (X-a)*(d-c)/(b-a)+c.
If you want to create a model with different ranges that bypasses this automatic range conversion there is a simple solution. You have to define the ranges of all variables to be the same and use the target functions to practically restrict the range to what you want. In the example above, change the range of X to [0,2] and it's target function to be min(1,old_target_function). In general, if the maximal range in the model is [a,b] and you want a range of [c,d], where a<=c<=d<=b you will have to change to use the range [a,b] and the target function min(d,max(c,old_target_function)).

The simple saving that we recommend you do frequently while you work is done by clicking the save icon, second from the top left. To name and save a model more precisely during, or at the end of a model sketch, click the model library icon at the top left. You will see that the model you have started working on is temporarily saved as ‘Default Model’. To change that, simply click this text in the model library (highlighting it) and rewrite the file name of your choice. (To save your work further, see the next question about exporting and importing XML files.)

In the model library, there are icons to export and import XML files. This is the best way to save and share models with your colleagues in a secure folder of your choice. Both export and import work like a typical ‘save as’ or ‘open’ function.

Once you have developed a comprehensive model library, you can skip between models by clicking them (to highlight), then using the open button.

Key F5 to refresh the page. Don’t forget to frequently save your work. Please report any bugs via our feedback page.

The drawing has an underlying dynamic semantics. Essentially, a drawing describes a system with multiple states with rules on how it changes from state to state. A state corresponds to an assignment to all the variables in the system. From a given state, the system transitions to a new state by updating all the variables simultaneously. In order to determine the new value for a variable the value of its target function is computed based on the values of other variables it is in relation with. Then, if the value of the target function is higher than the value of the variable, it increases by 1. If the value of the target function is lower than the value of the variable, it decreases by 1. If the value of the target function is the same as that of the variable, it does not change.

Bio Model Analyzer checks if the model becomes stable. That is, if regardless of the values the variables start with, there is one stability point that the model always reaches and then remains in. Bio Model Analyzer’s analysis is based on formal methods that were developed for the specification and verification of properties in concurrent software systems. The complexity of this analysis stems from the high number of possible states of such a system. A system with 10 variables, each ranging over 0-3, has 4^10 states, or roughly one million states. A system with 30 variables ranging over 0-1, has over one trillion states. Such large systems are impossible to manually inspect or visually inspect in a comprehendible layout. The formal analysis collects rules about which states of the system can be visited again and again. It uses this information to narrow down the possible states for the system to stabilize in. If the set of possible states to stabilize in collapses to just one state, then the analysis has been able to prove stabilization. If it does not, then the analysis tries to find either multiple sinks or several states that have a cycle between them.
We are also working on adding more types of analysis. We would appreciate your feedback.

Bio Model Analyzer is a model that abstracts the biological meanings that differentiate genes, transcription factors, proteins and the like. As the biologist, you need to identify your variables so they make sense to you (e.g. via their naming) but the actual analysis that Bio Model Analyzer provides does not account for differences between variables, beyond how they are named, connected, and given ranges and target functions.