Camila Costa

Ms. Camila Costa

orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
26 April 2022 in Forensic Science International: Genetics
Verified
Journal Article
Open access
CCCamila Costa
Camila Costa
CFCarolina Figueiredo
Carolina Figueiredo
Antonio AmorimAntonio Amorim
Antonio Amorim
SCSandra Costa
Sandra Costa
PFPaulo Miguel Ferreira
Paulo Miguel Ferreira
NPNádia Pinto
Nádia Pinto
To overcome the multifactorial complexity associated with the analysis and interpretation of the capillary electrophoresis results of forensic mixture samples, probabilistic genotyping methods have been developed and implemented as software, based on either qualitative or quantitative models. The former considers the electropherograms' qualitative information (detected alleles), whilst the latter also takes into account the associated quantitative information (height of allele peaks). Both models then quantify the genetic evidence through the computation of a likelihood ratio (LR), comparing the probabilities of the observations given two alternative and mutually exclusive hypotheses. In this study, the results obtained through the qualitative software LRmix Studio (v.2.1.3), and the quantitative ones: STRmix™ (v.2.7) and EuroForMix (v.3.4.0), were compared considering real casework samples. A set of 156 irreversibly anonymized sample pairs (GeneMapper files), obtained under the scope of former cases of the Portuguese Scientific Police Laboratory, Judiciary Police (LPC-PJ), were independently analyzed using each software. Sample pairs were composed by (i) a mixture profile with either two or three estimated contributors, and (ii) a single contributor profile associated. In most cases, information on 21 short tandem repeat (STR) autosomal markers were considered, and the majority of the single-source samples could not be a priori excluded as belonging to a contributor to the paired mixture sample. This inter-software analysis shows the differences between the probative values obtained through different qualitative and quantitative tools, for the same input samples. LR values computed in this work by quantitative tools showed to be generally higher than those obtained by the qualitative. Although the differences between the LR values computed by both quantitative software showed to be much smaller, STRmix™ generated LRs are generally higher than those from EuroForMix. As expected, mixtures with three estimated contributors showed generally lower LR values than those obtained for mixtures with two estimated contributors. Different software products are based on different approaches and mathematical or statistical models, which necessarily result in the computation of different LR values. The understanding by the forensic experts of the models and their differences among available software is therefore crucial. The better the expert understands the methodology, the better he/she will be able to support and/or explain the results in court or any other area of scrutiny.

20 January 2025 in Scientific Reports
Verified
Journal Article
Open access
PDF file
CCCamila Costa
Camila Costa
CFCarolina Figueiredo
Carolina Figueiredo
SCSandra Costa
Sandra Costa
PFPaulo Miguel Ferreira
Paulo Miguel Ferreira
Antonio AmorimAntonio Amorim
Antonio Amorim
NPNádia Pinto
Nádia Pinto
Technological advancements have allowed the detection of increasingly complex forensic genetics samples, as minimum amounts of DNA can now be detected in crime scenes or other settings of interest. The weight of the evidence depends on several parameters regarding the population and sample-related analytical factors, the latter in a greater number when the DNA amount is considered. This led to the development of probabilistic genotyping software (PGS), able to deal with the associated complexities. This study aims to evaluate the impact on the evidence’s weighing, when different analytical threshold values are used, and when different models and/or estimates for analytical artifacts, such as stutters or drop-in parameters, are considered. To reach this goal, three PGS, based on different statistical models, were used to analyze real casework pairs of samples composed of a mixture with either two or three estimated contributors, and a single-source sample associated. The obtained results show that the estimation of these parameters must not be overlooked, as they may considerably impact the outcome. This underlines the importance of proper parametrization in the analysis of forensic genetics identification problems when using complex samples, and the understanding by practitioners of how probabilistic genotyping informatics tools work to use them accurately.

8 September 2025 in Genes
Verified
Journal Article
Open access
PDF file
CCCamila Costa
Camila Costa
EPErica Pereira
Erica Pereira
SCSandra Costa
Sandra Costa
PFPaulo Miguel Ferreira
Paulo Miguel Ferreira
Antonio AmorimAntonio Amorim
Antonio Amorim
NPNádia Pinto
Nádia Pinto
Background: Probabilistic genotyping software has become an essential tool in forensic genetics, particularly for interpreting complex DNA mixtures. Previous studies measured the impact of considering widely divergent statistical approaches in quantifying evidence, both inter- and intra-software. At a much smaller scale, this data-driven study shows how different models implemented on distinct versions of the same tool may affect the results. Among the available tools, EuroForMix stands out as a quantitative, open-source software that models various aspects of the DNA profile, including artefacts like stutter peaks. Its freeware nature allowed the use of both versions 1.9.3. and 3.4.0, between which several updates were made, including the possibility to model both back and forward stutter, compared to only modeling back stutters inputted by the expert in the earlier version. Methods: A total of 156 real casework sample pairs (comprising mixtures with two or three estimated contributors and associated reference) from the Portuguese Scientific Police Laboratory were analyzed using both software versions. The same input data, containing alleles and artefactual peaks, were used to reflect operational conditions. Statistical measurements were compared and further investigated. Results: Most Likelihood Ratio values differed in less than one order of magnitude across versions. However, exceptions were found in more complex samples, such as those with more contributors, unbalanced contributions, or greater degradation. Conclusions: This work emphasizes the relevance of model selection in forensic evidence quantification, even when considering different versions of the same tool. The impact of different models in statistical evaluation depends on several factors, such as sample technical conditions, genotypic profiles, and population distribution.

28 October 2025 in Forensic Science International: Genetics
Verified
Journal Article
Open access
CCCamila Costa
Camila Costa
JÁJuan Carlos Álvarez
Juan Carlos Álvarez
SASofia Angeletti
Sofia Angeletti
SBSebastian Biagini
Sebastian Biagini
MCMariela Caputo
Mariela Caputo
NPNádia Pinto
Nádia Pinto
The increasing complexity of forensic genetic samples has driven the development and implementation of probabilistic genotyping software (PGS) to assist experts in quantifying the weight of evidence. A common challenge lies in quantifying the likelihood that a person of interest (PoI) is a contributor to a DNA mixture, especially in samples with low quantity and quality, stochastic effects (drop-in allele, dropout, and/or heterozygotic peaks imbalance), shared alleles, and stutter peaks. PGS can be based on either qualitative models (allele presence/absence) or quantitative (including allele peak heights) to assess the likelihood ratio, which consists of comparing the probability of observing the evidence given two hypotheses: "H1 = The PoI is a contributor to the evidence" and "H2 = The PoI is not a contributor nor genetically related to any contributor". Key parameters included in these calculations regarding population - allele frequency distributions, and co-ancestry coefficient -, analytical factors - drop-in, dropout, analytical threshold -, and stutter presence, can influence the outcomes depending on how they are set. This collaborative exercise, organized by the Spanish and Portuguese-speaking Working Group of the International Society for Forensic Genetics, aimed to assess the current state of knowledge, use, and implementation of PGS among forensic laboratories. The goal was to evaluate how non-binary informatics tools are applied in practice and to understand the methodologies used to statistically interpret complex DNA mixtures.Participants were asked to analyze 30 pairs of samples composed of DNA mixtures (with varying the number of contributors, mixture ratios, and degradation levels) and corresponding reference profiles, selected from the PROVEDIt database. Laboratories employed different tools and approaches to quantify the evidence's weight, including decisions regarding the number of contributors (NoC), population, laboratory, and case-specific parameters, such as coancestry coefficient, minimum allele frequency, dropout frequency, drop-in (frequency and modeling), analytical threshold, and the modeling of stutter peaks and degradation. Even though all laboratories received the same genotypic and frequency data, methodological differences led to different LR results, particularly for more complex samples with low-template and degraded DNA. The greatest differences were observed in the interplay between analytical thresholds and NoC estimation, with discrepancies amplified when alleles from minor contributors overlapped with expected stutter positions. This exercise highlights the importance of expert training and underscores the need for a comprehensive understanding of the statistical models underlying PGS. Ensuring accurate and consistent interpretation of complex DNA evidence requires not only technical proficiency but also an integrated approach to parameter selection and genotypic data evaluation.

6 January 2026 in Scientific Reports
Verified
Journal Article
Open access
PDF file
CCCamila Costa
Camila Costa
ECEduardo Conde-Sousa
Eduardo Conde-Sousa
CFCarolina Figueiredo
Carolina Figueiredo
SCSandra Costa
Sandra Costa
PFPaulo Miguel Ferreira
Paulo Miguel Ferreira
NPNádia Pinto
Nádia Pinto
The genetic similarity of unrelated individuals of the same population can be measured through the coancestry coefficient, which quantifies the probability of a pair of alleles randomly chosen having a common ancestor. It is thus expected that the higher this coefficient, the lower the genetic variation between individuals of a population. This may weaken the evidence’s weight in forensic genetics settings, whenever a person of interest cannot be excluded as a contributor to the evidence. Under simple assumptions, the impact of disregarding coancestry when quantifying the likelihood of an individual having contributed to a mixture evidence sample can be evaluated through an algebraic approach. This methodology becomes, however, unfeasible when considering all the information currently obtained from samples in forensic cases, whose analyses are software-dependent. Innovatively, this work assessed the impact of disregarding coancestry when quantifying the weight of the evidence in a modern population under the scope of identification problems involving mixtures. Methodology relied on both an algebraic approach (for simulated data) and an informatics one (for more complex real casework data). The results reinforce the importance of considering the proper population coancestry, whose impact is expected to be even higher for more isolated, substructured, or smaller populations.

9 May 2025 in International journal of legal medicine
Verified
Journal Article
Open access
PDF file
CCCamila Costa
Camila Costa
CFCarolina Figueiredo
Carolina Figueiredo
SCSandra Costa
Sandra Costa
PFPaulo Miguel Ferreira
Paulo Miguel Ferreira
AAAntónio Amorim
António Amorim
NPNádia Pinto
Nádia Pinto
Increasingly complex genetic samples are analyzed in forensic genetics routine, including mixtures to which more than one individual contributed. The standard problem relies on identification, aiming to quantify the likelihood of the donor of a reference sample being a contributor to the mixture. This is computed through a likelihood ratio (LR) and requires using devoted probabilistic genotyping software that may consider the quantity of the mixture’s DNA (quantitative tools), beyond only the presence/absence of specific alleles (qualitative tools). In any case, the mixture’s number of contributors (NoC) is a parameter that the user must introduce. Due to its nature, NoC is unknown for most real casework samples and needs to be estimated, which may be challenging due to poor DNA quality and quantity. This study aims to evaluate the impact of considering different NoC of real mixture samples (both over- and underestimating it after a first assessment of the expert) in identification problems through the pairwise comparison of LRs, using for the statistical assessment of both qualitative (LRmix Studio) and quantitative tools (EuroForMix and STRmix™). Different computational models showed different variations of the results, but for all, the impact was greater when considering a smaller NoC than the one initially estimated by the expert. Quantitative tools showed more sensitivity to NoC variation. Taking advantage of using real data, whose possible complexities surpass those of mock ones, this work highlights the impact that the NoC may have on the quantification of the proof, reinforcing the importance of its proper estimation.