Familias and Forensic genetics Thore. Egeland@umb Familias and Forensic genetics Thore.Egeland@umb.no Norwegian University of Life Sciences San Andrés Island, June 4, 2012 Presentation Leonor Petter Mostad, Daniel Kling Title: Famlias 1995: Pater 2011, Familias 2.0 (1.97) 2012 Code also released Open Familias (R) Forensic genetics: Statistical aspects emphasized; most of you know more of the technical, or lab aspects Kinship problems, not crime (mixtures) Your background? My backgroun

Purpose "..practical workshop about the statistical evaluation of forensic cases (paternity and other family relationships) by using Familias" Quoted from the mandate Key-words: Practical. Exercises. Focus on Familias Software. Case work Little/no mathematics Statistical evaluation: Useful also if you choose to report only on a verbal scale

Contents Principles Statistical evaluation Familias. Exercises LR (Paternity index) Non-standard cases Theta-correction Mutations Complex pedigrees Silent alleles. Familias. Exercises Course homepage arken.umb.no/~theg/familias/ user: familias pass word: andres Principles: Just a little bit. There is an underlying philosphy, science: ”Look to DNA”, i.e., other forensic disciplines are not as well founded Non-standard cases: Software is needed Exercises: Practical hands on exercises Homepage 9:00 - 10:00 Review of basic forensic genetics.  Principles. LR for simple cases. Spanish/Portuguese summary. 10:00 - 11:15 Introduction to Familias. Demo. Spanish/Portuguese summary 11:15 - 11:45 Coffee break 11:45 - 12:30 Familias exercises. Participants will use computer with Familias installed. 12:30 - 14:00 Lunch 14:00 - 16:00 Non-standard cases. Spanish/Portuguese summary. Complex relationships. Mutation. Theta-correction 16:00 - 17:00 Familias exercises. 17:00- 18:00 Discussion of exercises. Spanish/Portuguese summary. More questions, contributions from participants.

Review of basic forensic genetics Spanish/Portuguese summary Part I: 09-10 Review of basic forensic genetics Spanish/Portuguese summary Practicalities ended: Leonor: Review so far?

Books: Evett&Weir, Balding, Buckleton et al. Principles To evaluate the uncertainty of any given proposition it is necessary to consider at least one alternative proposition. Scientific interpretation is based on questions of the kind: “What is the probability of the evidence given the proposition?” Scientific evidence is conditioned not only by the competing propositions, but also by the framework of circumstances within which they are to be evaluated Why not only ”Man is the father”? Evaluation of evidence depends on alternative Alternative: ”unrelated”: Typically strong evidence Alternative: brother of AF: Typically weaker evidence Reporting officer: statements about DNA given proposition Court, Decision maker: statements about proposition given DNA What is the alternative? Need to know something about the case. ”All probabilities are conditional” Books: Evett&Weir, Balding, Buckleton et al. RMI

Exercise S1 Standard paternity case H1: The alleged father (AF) is the real father H2: AF and the child are unrelated. Most basic ideas can be illustrated by this simple example. There are alternatives to notation and formulation Explain notation

Exercise S1 a) Calculate LR. First marker Probabilities 0-1 (or 0-100%, not used in calculation) Conditional probabilities Explain LR pA close to 0: Strong evidence pA>>0: Weaker evidence 0.05 : From where? Interpretation: Numbers of verbal scale? Observe that we have used principles 1,2, and 3. Before next slide: Q: What is special , unique with DNA? Compared to say fingerprints A: Ther is lot of independent data Interpretation: “The data is 20 times more likely assuming AF to be the father compared to the alternative that some unknown man is the father”.

Exercise S1 c) Calculate LR. Second marker Only slightly more difficult Table, excel can cover all cases of genotype combinations.

Exercise S1 c) Calculate LR. Two markers Assumptions HW. Within marker. This is a bit hidden without more mathematics, but we have used that what is passed down from mother and father is independent and this inependence assumption is secured by HW LE. Linkage equilibrium No artefacts (mutation, silent alleles, genotyping errors) Interpretation: “The data is 200 times more likely assuming AF to be the father compared to the alternative that some unknown man is the father”. Assumptions?

Non-standard cases Complications Theta-correction Mutations Complex pedigrees Silent alleles Later

Part II: 10-1115: Introduction to Familias. Demo. Exercises S1-S3, S5. Plenum Videos available Spanish/Portuguese summary Leonor: review!i As videos 1-3; deliberately present mutations before we discuss in more detail

Part III: 1145-1230: Hands on Familias exercises Exercises S1-S6

Part IV: 14-16 Repetition. Exercises S4, S6. Plenum Non-standard cases

Non-standard cases Complications Theta-correction Mutations Complex pedigrees Silent alleles Need software How to deal with four major complications in Familias

Theta - correction Corrects for deviations from Hardy-Weinberg-Equilibrium What is HWE? Dependence between the two alleles Theta models this dependence Theta 0.01-0.03 Theta>0 : if allele A is seen many times , the probability of seing A again increases Longer story Hardy-Weinberg may not be valid for a number of reasons including non-random mating. At the simplestl evel the theta correction can be considered to correct for deviation from HW. Theta is between 0 and 1 and for 0 the usual results obtained assuming HW would apply. For theta>0 we see that the there is an excess of homozygotes (actually simple estimates of theta can be based on the difference betweenobserved and expected number of heterozygotes).

Revisit Exercise S1h (plenum) Only kinship, theta, parameter used in this window Revisit Exercise S1h (plenum)

Modelling mutations Mutation rate varies with Sex of parent and locus. Alleles tend to mutate to close alleles: Several models Assume a laboratory handles 1000 cases a year. With 10 markes and two meioses for each case, we would have something like 1000*10*2*0.005=100 mutations with an average mutation rate of 0,005. It’s hard to deal analytically with mutations and this was our original motviation for developing software. There are some ad-hoc mutations models around (at least I consider to to be ad hoc, since they tend not be based on a proper model.)

Mutations Revisit Exercise S2, S7 Females versus males Stable versus non-stable First two require two parameters Second two require only one Revisit Exercise S2, S7

Complex pedigrees I. Exercise S6

Complex pedigrees II. Exercise S6

Silent alleles I. Exercise S11 Silent alleles, null alleles

Silent alleles II. Familias 0.05+0.1+0.1+0.75=1

Silent alleles III. Exercises S11 (S12 theoretical)

Part V: 16-17 Hands on exercises: S9, S10, S11 If time: 'Extra' exercises on homepage: arken.umb.no/~theg/familias/

Part VI: 17-18 Discussion of exercises. Spanish/Portuguese summary. Questions? Exercises S1h, S4c, S5d, S9d S1h ”Discuss the assumptions underlying the calculations of this exercise: HWE; if not theta corrrection LD Mutations, silent alleles Other? Forensic markers are non-coding S4c: GF and GS share the same haplotype, freq 0.0025 LR=1/0.0025=400. Multiply 3.4? Problem: Y-based LR refers to different hypothesis; Could replace GF by say a brother and GS by say a brother S5d: W easier to interpret? Problem: prior. Why 50-50? S9d: Is there a best mutation model? Balance practical demands and theoretical demands Should a mutation model be used routinely for all markers? Case for YES: Should not adjust model once specifics of a case materialise

References Familias reference: Egeland, Mostad et al., see also http://familias.name/