Differential diagnosis in immunohistochemistry with Bayes theorem

Recommendations for immunocytochemistry in lung cancer typing: An update on a resource-efficient approach with large-scale comparative Bayesian analysis

OBJECTIVES

The majority of lung cancer cases are of advanced stage and diagnosis is usually made using minimally invasive small biopsies and cytological specimens. The WHO 2015 classification recommends limiting immunocytochemistry (ICC) to lung cancer typing and molecular testing drives for personalised therapies. An algorithm using Bayes' theorem could be useful for defining antibody profiles. This study aims to assess the impact of different antibody profiles for cytological samples on the accuracy of lung cancer typing with a large-scale Bayesian analysis.

METHODS

A retrospective examination of 3419 consecutive smears and/or cytospins diagnosed over 2011-2016 found 1960 primary lung cancer tumours: 972 adenocarcinomas (ADC), 256 squamous carcinomas (SQC), 268 neuroendocrine tumours (NET), and 464 non-small cell cancer-not otherwise specified (NSCC-NOS). The a priori and a posteriori probabilities, before and after ICC using antibodies singly or in combination, were calculated for different lung cancer types.

RESULTS

TTF-1 or CK7 alone improved the a posteriori probabilities of correct cytological typing for ADC to 86.5% and 95.8%, respectively. For SQC, using p40 (∆Np63) or CK5/6 together with CK5/14 led to comparable results (78.3% and 90.3%). With synaptophysin or CD56 alone, improvements in a posteriori probabilities to 87.5 and 90.3% for the correct recognition of NET could be achieved.

CONCLUSIONS

Based on morphological and clinical data, the use of two antibodies appears sufficient for reliable detection of the different lung cancer types. This applies to diagnoses that were finalised following ICC both on a clinical or cytological basis and on a histological basis.

Judging pathological assessment in cancer specimens

The pathologist plays a critical role in the multidisciplinary team in charge of treating cancer patients, as many of the therapeutic decisions rely on the information conveyed through the pathology reports. The task of the pathologist includes not only an accurate assessment of pathological T and N categories, but also the evaluation of other indicators of prognosis including quality of surgery, margins of resection, as well as additional histopathological and molecular markers that influence prognosis and could predict response to therapy.

Uncovering influence links in molecular knowledge networks to streamline personalized medicine

OBJECTIVES

We developed Resource Description Framework (RDF)-induced InfluGrams (RIIG) - an informatics formalism to uncover complex relationships among biomarker proteins and biological pathways using the biomedical knowledge bases. We demonstrate an application of RIIG in morphoproteomics, a theranostic technique aimed at comprehensive analysis of protein circuitries to design effective therapeutic strategies in personalized medicine setting.

METHODS

RIIG uses an RDF "mashup" knowledge base that integrates publicly available pathway and protein data with ontologies. To mine for RDF-induced Influence Links, RIIG introduces notions of RDF relevancy and RDF collider, which mimic conditional independence and "explaining away" mechanism in probabilistic systems. Using these notions and constraint-based structure learning algorithms, the formalism generates the morphoproteomic diagrams, which we call InfluGrams, for further analysis by experts.

RESULTS

RIIG was able to recover up to 90% of predefined influence links in a simulated environment using synthetic data and outperformed a naïve Monte Carlo sampling of random links. In clinical cases of Acute Lymphoblastic Leukemia (ALL) and Mesenchymal Chondrosarcoma, a significant level of concordance between the RIIG-generated and expert-built morphoproteomic diagrams was observed. In a clinical case of Squamous Cell Carcinoma, RIIG allowed selection of alternative therapeutic targets, the validity of which was supported by a systematic literature review. We have also illustrated an ability of RIIG to discover novel influence links in the general case of the ALL.

CONCLUSIONS

Applications of the RIIG formalism demonstrated its potential to uncover patient-specific complex relationships among biological entities to find effective drug targets in a personalized medicine setting. We conclude that RIIG provides an effective means not only to streamline morphoproteomic studies, but also to bridge curated biomedical knowledge and causal reasoning with the clinical data in general.

A pathologist-in-the-loop IHC antibody test selection using the entropy-based probabilistic method

Background:

Immunohistochemistry (IHC) is an important tool to identify and quantify expression of certain proteins (antigens) to gain insights into the molecular processes in a diseased tissue. However, it is a challenge for pathologists to remember the discriminative characteristics of the growing number of such antigens across multiple diseases. The complexity of their expression patterns, fueled by continuous discoveries in molecular pathology, gives rise to a combinatorial explosion that places an unprecedented burden on a practicing pathologist and therefore increases cost and variability of IHC studies.

Materials and Methods:

To tackle these issues, we have developed antibody test optimized selection method, a novel informatics tool to help pathologists in improving the IHC antibody selection process. The method uses extensions of Shannon's information entropies and Bayesian probabilities to dynamically build an efficient diagnostic tree.

Results:

A comparative analysis of our method with the expert and World Health Organization classification guidelines showed that the proposed method brings threefold reduction in number of antibody tests required to reach a diagnostic conclusion.

Conclusion:

The developed method can significantly streamline the antibody test selection process, decrease associated costs and reduce inter- and intrapathologist variability in IHC decision-making.

Primary lung cancer vs metastatic breast cancer: a probabilistic approach

In this study, a mathematical and probabilistic model is used to study the probability that a lung tumor is a primary vs a metastasis from cancer of the breast. The model uses information from immunohistochemical stains for thyroid transcription factor (TTF)-1, mammaglobin, p63, and estrogen receptor and epidemiologic data about primary lung and metastatic breast cancers in women. The results demonstrate that these 4 stains can yield nearly certain diagnoses in approximately 80% of tumors falling into the pool of this differential diagnosis. Nevertheless, uncertainty of diagnosis remains for the 19% of tumors in the pool that are negative for TTF-1, mammaglobin, and p63.