Machine Learning-Based Classification of β-Thalassemia Severity Levels

Classify β-Thalassemia severity levels based on hemoglobin variant patterns

β-Thalassemia (Mediterranean Anemia) is a genetic blood disorder caused by mutations in the HBB gene on chromosome 11, leading to reduced hemoglobin production. These mutations impair β-globin synthesis, resulting in anemia of varying severity. This study analyzes High-Performance Liquid Chromatography (HPLC) test results for hemoglobin variants (HbA,

HbA

, HbF) using a machine learning classification technique to categorize individuals as normal, β-Thalassemia minor, β-Thalassemia intermedia, or β-Thalassemia major.

Display the non-missing properties and corresponding values of the Thalassemia disease entity in a grid:

In[1]:=

thalassemiaProperties=

thalassemias

DISEASE

["Properties"];validProperties=SelectthalassemiaProperties,!MissingQ

thalassemias

DISEASE

[#]&;GridTransposevalidProperties,

thalassemias

DISEASE

/@validProperties,FrameAll

Out[1]=

average patient age

21.

average patient BMI

19.

kg/

average temperature

36.3093

°C

definition

an inherited blood disorder characterized by a decreased synthesis of one of the polypeptide chains that form hemoglobin; anemia results from this abnormal hemoglobin formation

average diastolic blood pressure

72.

mmHg

entity type list



disease



average patient height

1.1×

ICD-9 code

ICD-9 282.4

ICD-10 code

ICD-10 D56

name

thalassemias

prevalence

411400

people

prevalence rate

0.0054

medical specialty

{hematology}

average systolic blood pressure

1.3×

mmHg

treatment

{blood transfusions,iron chelation,folic acid supplements}

average patient weight

32.

Design a Graphics3D representation for the chromosome 11 and HBB genes[1]:

In[2]:=

chromosome11=Graphics3D

;gene[shift_,color1_,color2_]:=Graphics3D

;gene1=gene[2.5,Cyan,Magenta];gene2=gene[-2.5,Magenta,Cyan];Show[chromosome11,gene1,gene2,ImageSize400,BackgroundBlack,BoxedFalse,ViewPointFront]

Out[2]=

Create a Dataset to display hemoglobin patterns for normal and β-Thalassemia types[2]:

In[3]:=

hemoglobinPatterns=Dataset

;hemoglobinPatterns

Out[3]=

Hemoglobin Type	Normal	β-Thalassemia Minor	β-Thalassemia Intermedia	β-Thalassemia Major
HbA (%)	96-98	>88	10-50	0-30 (typically near 0)
HbA₂ (%)	2-3	>4	>4	>5
HbF (%)	<1	<5	10-50	Up to 95

Generate a visual gradient plot to interpret hemoglobin variant percentages, where

indicates normality and

indicates a higher level of β-Thalassemia:

In[4]:=

styledRow[label_,blendColors_]:=Row

;Column



Out[4]=

HbA (%)

HbA

(%)

HbF (%)

Prepare a training set by defining hemoglobin variant ranges for normal and β-Thalassemia types, mapping them to class labels:

In[5]:=

normalRanges=

;minorRanges=

;intermediaRanges=

;majorRanges=

;classes[className_,ranges_]:=Module{coordinates,classData},

CompoundExpression[

]

;normalClass=classes

;minorClass=classes

;intermediaClass=classes

;majorClass=classes

;individuals=Length/@{normalClass,minorClass,intermediaClass,majorClass};trainingSet=Join[normalClass,minorClass,intermediaClass,majorClass];features={"HbA","

HbA

","HbF"};trainingTable=Grid



Out[5]=

Class	Normal	β-Thalassemia Minor	β-Thalassemia Intermedia	β-Thalassemia Major	Total
Number of Individuals	1260	1000	1620	1680	5560

Train a machine to Classify hemoglobin data into β-Thalassemia status categories:

In[6]:=

thalassemiaClassifier=Classify[trainingSet,FeatureNamesfeatures];

Retrieve details of the β-Thalassemia classifier including number of examples, feature types, and class labels:

In[7]:=

Information[thalassemiaClassifier,"ExampleNumber"]

Out[7]=

5560

In[8]:=

Information[thalassemiaClassifier,"FeatureTypes"]

Out[8]=

HbANumerical,

HbA

Numerical,HbFNumerical

In[9]:=

Information[thalassemiaClassifier,"Classes"]

Out[9]=

{Normal,β-Thalassemia Intermedia,β-Thalassemia Major,β-Thalassemia Minor}

Use the β-Thalassemia classifier to determine classifications based on hemoglobin variant percentages:

In[10]:=

thalassemiaClassifier[{97.2,2.3,0.4}]

Out[10]=

Normal

In[11]:=

thalassemiaClassifier[{90.1,5,3}]

Out[11]=

β-Thalassemia Minor

In[12]:=

thalassemiaClassifier[{33.1,5,15}]

Out[12]=

β-Thalassemia Intermedia

In[13]:=

thalassemiaClassifier[{2.1,6,87}]

Out[13]=

β-Thalassemia Major

This work is dedicated to the memory of my dear siblings, Loran Masoud and Omar Masoud, victims of Beta Thalassemia, which claimed their lives. It is also dedicated to my only remaining sister, Lobna Masoud, who bravely battles this disease. Additionally, this work is dedicated to Resala Association in Egypt, a non-profit organization known for its remarkable efforts in charitable work, including combating β-Thalassemia. The association provides both financial and moral support to patients and ensures access to safe blood transfusion methods. I aspire to contribute to limiting the spread of this disease by studying it through data analysis and machine learning. For instance, exploring the relationship between genes and hemoglobin analysis results. Unfortunately, the required data is unavailable.

Source Metadata

Citation:
- [1] 23andMe.[Online]. Available: https://www.23andme.com/en-eu/topics/carrier/sickle-cell-anemia/[Accessed: January 22, 2025].
- [2] National Library of Medicine.[Online]. Available: https://www.ncbi.nlm.nih.gov/books/NBK1426/[Accessed: January 22, 2025].
- [3] MedlinePlus.[Online]. Available: https://medlineplus.gov/genetics/condition/beta-thalassemia/[Accessed: January 22, 2025].

Publisher Information

Contributed by: Akram Masoud

Machine Learning-Based Classification of β-Thalassemia Severity Levels

Source Metadata

Related Symbols

Publisher Information