Application of explainable artificial intelligence for healthcare: A systematic review of the last decade (2011-2022)

Machine learning in haematological malignancies

Digitalization of the medical record and integration of genomic methods into clinical practice have resulted in an unprecedented wealth of data. Machine learning is a subdomain of artificial intelligence that attempts to computationally extract meaningful insights from complex data structures. Applications of machine learning in haematological scenarios are steadily increasing. However, basic concepts are often unfamiliar to clinicians and investigators. The purpose of this review is to provide readers with tools to interpret and critically appraise machine learning literature. We begin with the elucidation of standard terminology and then review examples in haematology. Guidelines for designing and evaluating machine-learning studies are provided. Finally, we discuss limitations of the machine-learning approach.

Machine learning and artificial intelligence in haematology

The volume and complexity of scientific and clinical data in oncology have grown markedly over recent years, including but not limited to the realms of electronic health data, radiographic and histologic data, and genomics. This growth holds promise for a deeper understanding of malignancy and, accordingly, more personalized and effective oncologic care. Such goals require, however, the development of new methods to fully make use of the wealth of available data. Improvements in computer processing power and algorithm development have positioned machine learning, a branch of artificial intelligence, to play a prominent role in oncology research and practice. This review provides an overview of the basics of machine learning and highlights current progress and challenges in applying this technology to cancer diagnosis, prognosis, and treatment recommendations, including a discussion of current takeaways for clinicians.

https://ascopubs.org/doi/pdfdirect/10.1200/CCI.20.00049

Machine Learning in Oncology: What Should Clinicians Know?

/pdf/patients-with-cll-have-a-lower-risk-of-death-from-covid-19-3ay142dm.pdf

Patients with CLL have a lower risk of death from COVID-19 in the Omicron era

Infections have become the major cause of morbidity and mortality among patients with chronic lymphocytic leukemia (CLL) due to immune dysfunction and cytotoxic CLL treatment. Yet, predictive models for infection are missing. In this work, we develop the CLL Treatment-Infection Model (CLL-TIM) that identifies patients at risk of infection or CLL treatment within 2 years of diagnosis as validated on both internal and external cohorts. CLL-TIM is an ensemble algorithm composed of 28 machine learning algorithms based on data from 4,149 patients with CLL. The model is capable of dealing with heterogeneous data, including the high rates of missing data to be expected in the real-world setting, with a precision of 72% and a recall of 75%. To address concerns regarding the use of complex machine learning algorithms in the clinic, for each patient with CLL, CLL-TIM provides explainable predictions through uncertainty estimates and personalized risk factors.

/pdf/machine-learning-can-identify-newly-diagnosed-patients-with-3lls49d8od.pdf

Machine learning can identify newly diagnosed patients with CLL at high risk of infection.

Understanding of pandemics depends on the characterization of pathogen collections from well-defined and demographically diverse cohorts. Since its emergence in Congo almost a century ago, Human Immunodeficiency Virus Type 1 (HIV-1) has geographically spread and genetically diversified into distinct viral subtypes. Phylogenetic analysis can be used to reconstruct the ancestry of the virus to better understand the origin and distribution of subtypes. We sequenced two 3.6-kb amplicons of HIV-1 genomes from 3,197 participants in a clinical trial with consistent and uniform sampling at sites across 35 countries and analyzed our data with another 2,632 genomes that comprehensively reflect the HIV-1 genetic diversity. We used maximum likelihood phylogenetic analysis coupled with geographical information to infer the state of ancestors. The majority of our sequenced genomes (n = 2,501) were either pure subtypes (A-D, F, and G) or CRF01_AE. The diversity and distribution of subtypes across geographical regions differed; USA showed the most homogenous subtype population, whereas African samples were most diverse. We delineated transmission of the four most prevalent subtypes in our dataset (A, B, C, and CRF01_AE), and our results suggest both continuous and frequent transmission of HIV-1 over country borders, as well as single transmission events being the seed of endemic population expansions. Overall, we show that coupling of genetic and geographical information of HIV-1 can be used to understand the origin and spread of pandemic pathogens.

https://hal-pasteur.archives-ouvertes.fr/pasteur-03266546/document

Phylogenetic analysis of HIV-1 shows frequent cross-country transmission and local population expansions.

Abstract Background HIV-1 set-point viral load (SPVL) is predictive of disease progression and shows variability across HIV-1-positive (HIV+) persons. Various factors may influence SPVL including viral features, environmental exposure and host genetics. To identify single nucleotide polymorphisms (SNPs) associated with SPVL, we performed a genome-wide association study (GWAS) on a subset of participants from the Strategic Timing of AntiRetroviral Treatment (START) study covering a demographically diverse population. Methods. Consenting participants were antiretroviral therapy (ART)-naïve and SPVL was taken as log10(HIV RNA) at study entry. Genotypic data were generated on a custom content Affymetrix Axiom SNP array covering 770,558 probes. The Ensembl Gene database, assembly GRCh37.p13, was used for annotation. Principal component analysis (PCA) was used to identify population structures, and analysis of variance (ANOVA) was performed to detect associations between SNPs and SPVL. SNPs with zero variance or minor allele frequency (MAF) ≤0.05 were removed. Results. Among the 2,544 participants, PCA showed distinct population structures with strong separation between black (n = 578) and nonblack (n = 1,966) participants, Figure 1. ANOVA was performed independently on both subsets. Two SNPs located in the Major Histocompatibility Complex (MHC) class I region of chromosome six reached genome-wide significance (P < 5 × 10–8) in the non-black population: rs4418214 (P = 1.74 × 10-10), and rs57989216 (P = 3.96 × 10–8), Figure 2. Two additional SNPs, rs9264942 (P = 5.99 × 10–8) and rs7356880 (P = 9.69 × 10–8), in the same region approached significance. The minor alleles of all four SNPs were associated with lower SPVL, Figure 3. While no SNPs reached genome-wide significance in the black group, we observed similar trends toward lower SPVL for both rs4418214 and rs57989216. Conclusion. In this study we confirm the association of a previously reported SNP (rs4418214) and identify a novel candidate SNP (rs57989216) associated with lower SPVL in a population of nonblack, ART-naïve HIV+ persons. Current findings suggest that the effects of these SNPs are consistent across race groups, but further studies are required to confirm this. Our results support previous findings that variation in the MHC class I region is a major host determinant of HIV-1 control. Disclosures D. D. Murray, Centre of Excellence for Health, Immunity and Infectious diseases (CHIP), Department of Infectious Diseases, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark: Employee, Salary. J. M. Molina, Gilead: Scientific Advisor, Consulting fee. Merck: Scientific Advisor, Consulting fee. ViiV: Scientific Advisor, Consulting fee. Teva: Scientific Advisor, Consulting fee.

/pdf/1284-study-of-single-nucleotide-polymorphisms-associated-3j7fsx1676.pdf

1284. Study of Single Nucleotide Polymorphisms Associated with HIV-1 Set-Point Viral Load in Antiretroviral Therapy-Naïve HIV-Positive Participants of the START Study

Man-Hung Eric Tang

Papers

Machine learning can identify newly diagnosed patients with CLL at high risk of infection.

Phylogenetic analysis of HIV-1 shows frequent cross-country transmission and local population expansions.

1284. Study of Single Nucleotide Polymorphisms Associated with HIV-1 Set-Point Viral Load in Antiretroviral Therapy-Naïve HIV-Positive Participants of the START Study