Using Mendelator

This is an app useable on all mobiles, tablets and desktops that calculates the probabilities of different inheritance patterns in a family whose members are affected by medical condition.

The app has the following features:


 

Sign Up

Signing up requires submitting a unique user name and email address. Also required is an activation key code supplied offline by the Site Admin. An email is sent to the submitted email address with an activation link. Clicking this link activates the account. Lost and retrieved passwords are sent to the submitted email account.

All comments are visible to all logged-in users. Users' previous comments are visible to other logged-in users by clicking the Users tab but not editable. The Site Admin can remove a user or edit cases at his discretion.

Users who are not logged in can still use this Calculator. However, they cannot save their settings, submit comments, or access the full help page.

 

Home Page

If the user is not logged in, this brings up the sign-up page for creating a new user account. To log into an existing account, click 'Log in' to the right of the top menu bar.

If the user is logged in, the Home page shows a live feed of recent comments from all users, and a form to send a new comment. If desired, the user can upload a picture with the comment. To the top left of the screen is the user's name. To view the comments of a particular user, click the Users tab on the top menu bar, then select the user from the list.

 

Phenotype

Some information on conditions typically presenting with the observed phenotype is required for inheritance calculations.

First, since the likelihood of affected individuals depends depends greatly on the disease penetrance (the likelihood of a dominant or recessive mutation actually manifesting with one or both disease alleles respectively), the typical penetrance of the inherited diseases that can present with the observed phenotype must be entered. Note that entering 100% may yield strange results because here if a single definite carrier individual does not manifest, the pattern is excluded. For example, in most inheritance patterns, one parent and one grandparent would have to be a definite carrier for the case to be affected, and have to be affected themselves if penetrance is 100%. For a more realistic high penetrance estimate, consider entering 99%. (This is why the default penetrance setting is at 99% not 100%.)

The overall prevalence of the phenotype will determine the likelihood of chance occurrences in the family. If a phenotype is common, one may expect even sporadic diseases to have more than one family member affected. Enter the typical lifetime prevalence level of observed phenotype.

 

Preset Phenotypes

To aid in entering data on penetrance, prevalence and relative frequencies of different inheritance patterns, the user may select a preset profile from a list of common situations where there is a mix of different inheritance patterns resulting in a particular phenotype.

Any suggestions on adding data for other phenotypes are welcome!

 

Relatives (Affected or Unaffected)

Enter the number of each type of relative. Note that sensitivity and specificity calculations just pool the numbers of male and female cousins rather than assigning each cousin to a particular aunt or uncle.

This form to calculate specificity and sensitivity does not provide the matching inheritance pattern; the latter can be inferred from the Inheritance Probabilities form or may be already known. Sensitivity and Specificity depend only on the family size and make-up. Therefore there are no entries for affected versus unaffected family members. Check the results row for the inheritance pattern determined on the Inheritance Probabilities form or the pattern in which you are interested. The Sensitivity and Specificity apply to how strongly the family can confirm the inheritance according to the criteria, and depend on the family size and make-up, not on which family members are actually affected. The family make-up is like a sample size; for example, if we already know a test is positive (i.e. matching the criteria for a certain inheritance pattern), we do not need to check this again. Instead, we want to determine how likely it was that we got this positive result based on the sample size. A greater number of eligible relatives for a particular inheritance pattern (e.g. sibs for AR inheritance) will increase the sensitivity of that family for getting a positive result for that pattern, while a greater number of ineligible relatives (e.g. non sibs for AR inheritance) will increase the specificity for correctly identifying some pattern other than AR, i.e make a false positive for AR less likely.

 

Affected Relatives

Set each grandparent and parent to affected or unaffected, and use the buttons to add further relatives to build up the complete family. If the case is older and there are children or nephews or neices affected, consider one of the younger generation to be the case, and set the actual case to be an affected parent or uncle or aunt. This will better represent the three generations tested by this App.

 

Relative Inheritance Frequencies

For a particular disease phenotype, in the general population some inheritance patterns will be more common than others. The default values, where sporadic occurrence is most likely at 80%, followed by familial inheritance at 10%, and Mendelian patterns make up the rest, can be changed. For example, in clinical young onset parkinsonism it might be considered that AD or AR inheritance is actually more likely than sporadic occurrence. Obviously the combined values must still add up to 100%.

 

Inheritance Criteria

The probability calculations determine the relative chance that a particular pattern of affected and unaffected relatives reflects a particular inheritance pattern. But it is also helpful to have a positive or negative "test" result. Hence mutually exclusive criteria are set so that each family of affected and unaffected relatives satisfies one, and only one, inheritance pattern. The criteria were set based on inheritance rules and using calculated probability results in different families to strike the best overall balance between sensitivity and specificity.

 

Unadjusted Probabilities

These are the probabilities that the entered affected and unaffected relatives in the family reflect a certain inheritance pattern, assuming that the inheritance patterns are initially equally likely in the general population, i.e. have equal relative frequencies.

 

Adjusted Probabilities

These are the probabilities that the entered pattern of affected and unaffected relatives reflects a certain inheritance pattern, adjusting for the fact that some inheritance patterns more commonly produce the observed phenotype than other inheritance patterns, as set by the Relative Frequencies table.

 

Positive Predictive Value/ False Discovery Rates

Given the critera test is positive for one inheritance type, the Positive Predictive Value (PPV) is the relative chance that this is a true positive result, i.e. true positive cases (true criteria matches) divided by all positive cases (true and false criteria matches).

Since it is already given that there is an inheritance match (the denominator of the PPV formula), the PPV is not measuring the same thing as probability, but complements it and is something more akin to specificity. For a more formal measure of the "accuracy" of the criteria match, switch to the Sensitivity and Specificity page, and read off the values for the matched pattern in that family.

Additional affected or unaffected candidate members will make the Adjusted Probability more or less likely compared to other patterns, but will have a different (or sometimes no) effect on PPV because the latter depends not on the probability of the true pattern (it just has to be enough to satisfy the criteria) but on the probability that other patterns do not mimic the true pattern.

It therefore follows that once an inheritance pattern is satisfied, while the presence of further candidate relatives will affect the Positive Predictive Value because they may affect the likelihood of other patterns mimicking the true pattern, the Predictive Value does not vary according to whether or not these additional relatives are actually affected, as long as they do not switch to matching an alternative pattern. So, for example, if there are three siblings, and no other relatives, having the third sibling affected actually reduces the probability of AR compared to AD or sometimes Maternal inheritance, because the 100% penetrance chance for each sibling is only 25% versus 50% (AD) or 100% (Maternal). It does not affect the Positive Predictive Value for AR, because the criteria already matched, and false positives depend only on the family make-up (i.e "sample size") influencing the potential for a confounding result.

The False Discovery Rates are the corresponding values for the inheritance patterns that do not match the criteria, i.e. the false positives (false criteria matches) for this inheritance divided by all positives (true and false criteria matches) for the true pattern. All the false discovery rates together with the positive predictive value will add up to 100%. False positives happens when two or more inheritance types could account for a certain family of affected relatives, but the criteria must be mutually exclusive. For example, an affected maternal grandmother and mother could occur in AD or XD inheritance, if the case was female, but it was empirically determined that this pattern is more "specific" for XD inheritance.

 

Negative Predictive Value

The Negative Predictive Value is the % proportion that a negative test result is a true negative for a certain inheritance, i.e. the true negatives for a certain inheritance (true criteria non match), divided by all negatives for that inheritance (true and false criteria non matches). The negative predictive value is similar to the specificity: the numerators are the same but the denominator for specificity is the total probability of not being this inheritance pattern (true negatives + false positives) instead of the total probability of not matching the criteria for this inheritance pattern (true negatives + false negatives). Unlike sensitivity, both of these values depend on the relative population frequencies of the different inheritance patterns.

Note that while the probability columns depend on the actual relatives affected, and the PPV/FDR column depends on the criteria match, the NPV column depends on neither. It is a general calculation of true negative rates for that family before considering which relatives are affected, and gives a general indication of how well the family would correctly excude an inheritance pattern if the criteria match was not for that pattern.

While the FDR is an indication of the predictive error when the criteria match a particular alternative inheritance pattern, the NPV is an indication of true negatives when the criteria match was for any pattern other than the one in question. That is why the NPV is shown even for the pattern where there is a criteria match; it indicates the chance of a true negative for that pattern, if the criteria match in that family had happened to be any other pattern.

 

Calculation of Family History Sensitivity & Specificity

This page will calculate, given a certain family make-up and disease penetrance, the sensitivity, specificity and positive and negative likelihood ratios for a certain inheritance actually appearing as that inheritance in a family.

Sensitivity answers the question, "If I suspect that an individual has a disease of a certain inheritance (including sporadic), what is the chance that the family is large enough to reveal that inheritance according to the criteria?"

Specificity answers the question, "In considering a particular mode of inheritance, when the pattern of affected relatives in a family does not match that inheritance (perhaps because of the probability findings on the other page), what is the probability that the underlying inheritance is truly not this particular inheritance?"

+ve and -ve Likelihood Ratios answer the question, "If I suspect an individual has a certain disease phenotype, how much more or less likely are they to have that disease if their family history is positive or negative (on the probability page) for the mode of inheritance of the disease?". In other words, the family history is considered a positive or negative "test" for that disease, one that can be combined with other tests to build an overall picture of the likelihood of that diagnosis.

 

Inheritance

For Sensitivity, Specificity and Matthews Correlation Coefficient, this column refers to the underlying actual inheritance pattern in the family. The three measures are showing how good the family make-up is at identifying this underlying inheritance pattern. So Sensitivity, Specicity and MCC are read as, "Given a certain underlying inheritance pattern, how good would the family be at corretly identifying this pattern?"

For +ve and -ve Likelihood Ratios, the column refers to the inheritance pattern whose criteria are matched by the pattern of affected and unaffected members. In other words this is the test result, not the underlying pattern. So the likelihood ratios are read as, "Given a certain inheritance criteria match, how much more or less likely is the underlying inheritance pattern to be this matching pattern?"

 

Sensitivity

Sensitivity is the probabililty that if the family has a disease that has a certain mode of inheritance, the family would actually show this pattern according to the criteria. Because it is given that the family truly has the inheritance type in question, the relative frequencies of inheritance modes for the phenotype does not affect the sensitivity, nor does the the actual pattern of who is affected. In other words, we already know the underlying inheritance mode; the question is how likely can a certain size family reveal it. The sensitivity is like a test of sample size; it tests the likelihood that the number and type of relatives in the family would mean there would be sufficient affected members to satisfy the criteria for that inheritance mode.

 

Specificity

Specificity is the probabililty that if the family did not have an inheritance mode, the family would correctly be identified as such by the criteria. Like sensitivity, specificity does not depend on affected versus unaffected relatives, but it does depend on the relative inheritance frequencies; it is only given that the true inheritance is not the inheritance mode in question, so if certain inheritances are more frequent in the first place, they will have a greater weighting on the likelihood of error. For example, if underlying Autosomal Dominant inheritance is the main type falsely identified as X-linked Dominant, then the Specificity of that family as a test for X-Linked Dominant is lower if Autosomal Dominant inheritance is relatively common.

If the penetrance is nearly 100%, specificity values often seem very high. This is because with carriers guaranteed to manifest, often only a few inheritance patterns may mimic others, for example AD with affected maternal grandmother and mother mimicking XD. If sporadic prevalence is low, it is very unlikely that this lack of inheritance will randomly produce all the correct carriers and only the correct carriers.

 

Matthews Correlation Coefficient (MCC)

This is a measure of how good a test is, considering both sensitivity and specificity, in this case how good the family history is at accurately revealing the presence or absence of different inheritance patterns. It is a balanced measure that can be used even if the inheritance patterns have widely varying relative frequencies, or have widely varying likelihoods in the family, and is better than accuracy or the area under the Receiver Operating Characteristic curve, for example. A value of 1 is a perfect test, zero a test with random results, and -1 a test that is always wrong.

When there are widely different relative frequencies, sensitivity and specificity may both be very high, but this may belie a major false positive rate compared to true positive rate. For example, consider 100 cases where there is only about 1 in 100 actual positive cases. Let us suppose the test is 99% sensitive, so the true positive rate is 0.99 cases out of 100, and just 0.01 false negative. However let us also suppose that there are twice as many false positives as true positives (FP = 2), and the remainder are true negatives (TN = 97). So despite the fact that a positive result is twice as likely be wrong as right, the specificity looks high TN / (TN+FP) = 98%, but this is only because so few cases are positive compared to the total number of cases. The MCC correctly reveals the suboptimal nature of the test, with a value of around 0.57 out of a perfect 1.

 

Likelihood Ratio +ve

The positive likelihood ratio measures how much more likely is a certain mode of inheritance if the family history actually shows that mode and is calculated from sensitivity/ (1 - specificity), ie true positives over false positives. For example, a value of 2 indicates that if, before looking at the family history, there was a certain probability that the disease had a certain inheritance, a positive family history would double this probabilitiy.

An odds ratio is something different. This would be the fraction of cases with a certain underlying inheritance mode showing that pattern divided by those that didn't, divided by the fraction of cases without the underlying inheritance mde who nevertheless showed the pattern divided by those that didn't.

The relative risk is the proportion of cases with an underlying mode inheritance who showed it out of all the cases with the underlying inheritance, divided by the proportion of cases without the underlying mode of inheritance but who nevertheless showed it out of all the cases without the underlying inheritance pattern.

 

Likelihood Ratio -ve

The negative likelihood ratio measures how much less likely is a certain inheritance pattern if the family history fails to show that pattern and is calculated from (1 - sensitivity) / (specificity), ie false negatives over true negatives. A value of 0.5 indicates that if, before looking at the family history, there was a certain probability that the disease had a certain inheritance, a negative family history would halve this probabilitiy.

 

Calculation of Inheritance Probabilities

Users who are not logged in can still use this Calculator. However, they cannot save their settings, submit comments, or access the full help page.

This function will calculate the relative probabilities of different inheritances when there is a particular pattern of similarly affected or unaffected members in a patient's family.

It answers the question, "Given certain affected and unaffected family members, and a certain suspected disease penetrance or prevalence, what is the relative likelihood of different modes of inheritance?".

It also determines if the family pattern matches mutually exclusive criteria for each mode of inheritance and the positive and negative predictive values for meeting these criteria depending on the number and type of family members.

 

Version Number

This is version 0.3 of the Mendelian Calculator application.