Avian influenza

From Wikipedia, the free encyclopedia
https://en.wikipedia.org/wiki/Avian_influenza

Avian influenza, also known as avian flu or bird flu, is a disease caused by the influenza A virus, which primarily affects birds but can sometimes affect mammals including humans. Wild aquatic birds are the primary host of the influenza A virus, which is enzootic (continually present) in many bird populations.

Symptoms of avian influenza vary according to both the strain of virus underlying the infection, and on the species of bird or mammal affected. Classification of a virus strain as either low pathogenic avian influenza (LPAI) or high pathogenic avian influenza (HPAI) is based on the severity of symptoms in domestic chickens and does not predict severity of symptoms in other species. Chickens infected with LPAI display mild symptoms or are asymptomatic, whereas HPAI causes serious breathing difficulties, significant drop in egg production, and sudden death. Domestic poultry may potentially be protected from specific strains of the virus by vaccination.

Humans and other mammals can only become infected with avian influenza after prolonged close contact with infected birds. In mammals including humans, infection with avian influenza (whether LPAI or HPAI) is rare. Symptoms of infection vary from mild to severe, including fever, diarrhea, and cough.

Influenza A virus is shed in the saliva, mucus, and feces of infected birds; other infected animals may shed bird flu viruses in respiratory secretions and other body fluids (e.g., cow milk). The virus can spread rapidly through poultry flocks and among wild birds. A particularly virulent strain, influenza A virus subtype H5N1 (A/H5N1) has the potential to decimate domesticated poultry stocks and an estimated half a billion farmed birds have been slaughtered in efforts to contain the virus.

Highly pathogenic avian influenza

Because of the impact of avian influenza on economically important chicken farms, a classification system was devised in 1981 which divided avian virus strains as either highly pathogenic (and therefore potentially requiring vigorous control measures) or low pathogenic. The test for this is based solely on the effect on chickens – a virus strain is highly pathogenic avian influenza (HPAI) if 75% or more of chickens die after being deliberately infected with it. The alternative classification is low pathogenic avian influenza (LPAI). This classification system has since been modified to take into account the structure of the virus' haemagglutinin protein. Other species of birds, especially water birds, can become infected with HPAI virus without experiencing severe symptoms and can spread the infection over large distances; the exact symptoms depend on the species of bird and the strain of virus. Classification of an avian virus strain as HPAI or LPAI does not predict how serious the disease might be if it infects humans or other mammals.

Since 2006, the World Organization for Animal Health requires all LPAI H5 and H7 detections to be reported because of their potential to mutate into highly pathogenic strains.

Virology

Main article: Influenza A virus

A transmission electron micrograph (TEM) of the reconstructed 1918 pandemic influenza virus. The bottom structure represents membrane debris from the cells used to amplify the virus.

Avian influenza is caused by the influenza A virus which principally affects birds but can also infect humans and other mammals. Influenza A is an RNA virus with a genome comprising a negative-sense, RNA segmented genome that encodes for 11 viral genes. The virus particle (also called the virion) is 80–120 nanometers in diameter and elliptical or filamentous in shape. There is evidence that the virus can survive for long periods in freshwater after being excreted in feces by its avian host, and can withstand prolonged freezing.

There are two proteins on the surface of the viral envelope; hemagglutinin and neuraminidase. These are the major antigens of the virus against which neutralizing antibodies are produced. Influenza virus epidemics and epizootics are associated with changes in their antigenic structure.

Hemagglutinin (H) is an antigenic glycoprotein which allows the virus to bind to and enter the host cell. Neuraminidase (N) is an antigenic glycosylated enzyme which facilitates the release of progeny viruses from infected cells. There are 18 known types of hemagglutinin, of which H1 thru H16 have been found in birds, and 11 types of neuraminidase.

Subtypes

Subtypes of influenza A are defined by the combination of H and N proteins in the viral envelope; for example, "H5N1" designates an influenza A subtype that has a type-5 hemagglutinin (H) protein and a type-1 neuraminidase (N) protein. The subtyping scheme only takes into account the two envelope proteins, not the other proteins coded by the virus' RNA. Almost all possible combinations of H (1 thru 16) and N (1 thru 11) have been isolated from wild birds. Further variations exist within the subtypes and can lead to very significant differences in the virus's ability to infect and cause disease.

Influenza virus nomenclature

Diagram of influenza nomenclature

To unambiguously describe a specific isolate of virus, researchers use the internationally accepted Influenza virus nomenclature, which describes, among other things, the species of animal from which the virus was isolated, and the place and year of collection. As an example, A/chicken/Nakorn-Patom/Thailand/CU-K2/04(H5N1):

• A stands for the genus of influenza (A, B or C)
• chicken is the animal species the isolate was found in (note: human isolates lack this component term and are thus identified as human isolates by default)
• Nakorn-Patom/Thailand is the place this specific virus was isolated
• CU-K2 is the laboratory reference number that identifies it from other influenza viruses isolated at the same place and year
• 04 represents the year of isolation 2004
• H5 stands for the fifth of several known types of the protein hemagglutinin
• N1 stands for the first of several known types of the protein neuraminidase.

Other examples include: A/duck/Hong Kong/308/78(H5N3), A/avian/NY/01(H5N2), A/chicken/Mexico/31381-3/94(H5N2), and A/shoveler/Egypt/03(H5N2).

Genetic characterization

Analysis of the virus' genome enables researchers to determine the order of its nucleotides. Comparison of the genome of a virus with that of a different virus can reveal differences between the two viruses. Genetic variations are important because they can change amino acids that make up the influenza virus’ proteins, resulting in structural changes to the proteins, and thereby altering properties of the virus. Some of these properties include the ability to evade immunity and the ability to cause severe disease.

Genetic sequencing enables influenza strains to be further characterised by their clade or subclade, revealing links between different samples of virus and tracing the evolution of the virus over time.

Species barrier

Humans can become infected by the avian flu if they are in close contact with infected birds. Symptoms vary from mild to severe (including death), but as of December 2024 there have been no observed instances of sustained human-human transmission.

There are a number of factors that generally prevent avian influenza viruses from causing epidemics in humans or other mammals.

• The viral HA protein of avian influenza binds to alpha-2,3 sialic acid receptors, which are present in the respiratory tract and intestines of avian species, while human influenza HA binds to alpha-2,6 sialic acid receptors, which are present in the human upper respiratory tract.
• The myxovirus resistance protein (Mx1) is an important antiviral restriction factor that inhibits the replication of avian influenza viruses in particular. Human-adapted strains of IAV display reduced sensitivity to human Mx1 compared with avian strains.
• Other factors include the ability to replicate the viral RNA genome within the host cell nucleus, and to transmit between individuals.

Influenza viruses are constantly changing as small genetic mutations accumulate, a process known as antigenic drift. Over time, mutation may lead to a change in antigenic properties such that host antibodies (acquired through vaccination or prior infection) do not provide effective protection, causing a fresh outbreak of disease.

The segmented genome of influenza viruses facilitates genetic reassortment. This can occur if a host is infected simultaneously with two different strains of influenza virus; then it is possible for the viruses to interchange genetic material as they reproduce in the host cells. Thus, an avian influenza virus can acquire characteristics, such as the ability to infect humans, from a different virus strain. The presence of both alpha 2,3 and alpha 2,6 sialic acid receptors in pig tissues allows for co-infection by avian influenza and human influenza viruses. This susceptibility makes pigs a potential "melting pot" for the reassortment of influenza A viruses.

Epidemiology

History

Avian influenza (historically known as fowl plague) is caused by bird-adapted strains of the influenza type A virus. The disease was first identified by Edoardo Perroncito in 1878 when it was differentiated from other diseases that caused high mortality rates in birds; in 1955 it was established that the fowl plague virus was closely related to human influenza. In 1972, it became evident that many subtypes of avian flu were endemic in wild bird populations.

Between 1959 and 1995, there were 15 recorded outbreaks of highly pathogenic avian influenza (HPAI) in poultry, with losses varying from a few birds on a single farm to many millions. Between 1996 and 2008, HPAI outbreaks in poultry have been recorded at least 11 times and 4 of these outbreaks have resulted in the death or culling of millions of birds. Since then, several virus strains (both LPAI and HPAI) have become endemic among wild birds with increasingly frequent outbreaks among domestic poultry, especially of the H5 and H7 subtypes.

Transmission and prevention

The eight major flyways used by shorebirds (waders) on migration

  Pacific

  Mississippi and Amazon

  West Atlantic

  East Atlantic

  Mediterranean and Black Sea

  West Asia and Africa

  Central Asia and India

  East Asia and Australasia

Birds – Influenza A viruses of various subtypes have a large reservoir in wild waterbirds of the orders Anseriformes (for example, ducks, geese, and swans) and Charadriiformes (for example, gulls, terns, and waders) which can infect the respiratory and gastrointestinal tract without affecting the health of the host. They can then be carried by the bird over large distances, especially during annual migration. Infected birds can shed avian influenza A viruses in their saliva, nasal secretions, and feces; susceptible birds become infected when they have contact with the virus as it is shed by infected birds. The virus can survive for long periods in water and at low temperatures, and can be spread from one farm to another on farm equipment. Domesticated birds (chickens, turkeys, ducks, etc.) may become infected with avian influenza A viruses through direct contact with infected waterfowl or other infected poultry, or through contact with contaminated feces or surfaces.

Avian influenza outbreaks in domesticated birds are of concern for several reasons. There is potential for low pathogenic avian influenza viruses (LPAI) to evolve into strains which are high pathogenic to poultry (HPAI), and subsequent potential for significant illness and death among poultry during outbreaks. Because of this, international regulations state that any detection of H5 or H7 subtypes (regardless of their pathogenicity) must be notified to the appropriate authority. It is also possible that avian influenza viruses could be transmitted to humans and other animals which have been exposed to infected birds, causing infection with unpredictable but sometimes fatal consequences.

When an HPAI infection is detected in poultry, it is normal to cull infected animals and those nearby in an effort to rapidly contain, control and eradicate the disease. This is done together with movement restrictions, improved hygiene and biosecurity, and enhanced surveillance. 

Humans – Avian flu viruses, both HPAI and LPAI, can infect humans who are in close, unprotected contact with infected poultry. Incidents of cross-species transmission are rare, with symptoms ranging in severity from no symptoms or mild illness, to severe disease that resulted in death. As of February, 2024 there have been very few instances of human-to-human transmission, and each outbreak has been limited to a few people. All subtypes of avian Influenza A have potential to cross the species barrier, with H5N1 and H7N9 considered the biggest threats.

In order to avoid infection, the general public are advised to avoid contact with sick birds or potentially contaminated material such as carcasses or feces. People working with birds, such as conservationists or poultry workers, are advised to wear appropriate personal protection equipment.

Other animals – a wide range of other animals have been affected by avian flu, generally due to eating birds which had been infected. There have been instances where transmission of the disease between mammals, including seals and cows, may have occurred.

Pandemic potential

Influenza viruses have a relatively high mutation rate that is characteristic of RNA viruses. The segmentation of the influenza A virus genome facilitates genetic recombination by segment reassortment in hosts who become infected with two different strains of influenza viruses at the same time. With reassortment between strains, an avian strain which does not affect humans may acquire characteristics from a different strain which enable it to infect and pass between humans – a zoonotic event. It is thought that all influenza A viruses causing outbreaks or pandemics among humans since the 1900s originated from strains circulating in wild aquatic birds through reassortment with other influenza strains. It is possible (though not certain) that pigs may act as an intermediate host for reassortment.

As of June 2024, there is concern about two subtypes of avian influenza which are circulating in wild bird populations worldwide, H5N1 and H7N9. Both of these have potential to devastate poultry stocks, and both have jumped to humans with relatively high case fatality rates.

Surveillance

The Global Influenza Surveillance and Response System (GISRS) is a global network of laboratories that monitor the spread of influenza with the aim to provide the World Health Organization with influenza control information and to inform vaccine development. Several millions of specimens are tested by the GISRS network annually through a network of laboratories in 127 countries. As well as human viruses, GISRS monitors avian, swine, and other potentially zoonotic influenza viruses.

Vaccine

Poultry – it is possible to vaccinate poultry against specific strains of HPAI influenza. Vaccination should be combined with other control measures such as infection monitoring, early detection and biosecurity.

Humans – Several "candidate vaccines" are available in case an avian virus acquires the ability to infect and transmit among humans. There are strategic stockpiles of vaccines against the H5N1 subtype, which is considered the biggest risk. A vaccine against the H7N9 subtype, which has also infected humans, has undergone a limited amount of testing. In the event of an outbreak, the "candidate" vaccine would be rapidly tested for safety as well as efficacy against the zoonotic strain, and then authorised and distributed to vaccine manufacturers.

Zoonotic influenza vaccine Seqirus is authorized for use in the European Union. It is an H5N8 vaccine that is intended to provide acquired immunity against H5 subtype influenza A viruses.

Influenza A virus subtype H5N1

Influenza A subtype H5N1

Further information: Influenza A virus subtype H5N1 and Transmission and infection of H5N1

The highly pathogenic influenza A virus subtype H5N1 is an emerging avian influenza virus that is causing global concern as a potential pandemic threat. It is often referred to simply as "bird flu" or "avian influenza", even though it is only one of many subtypes.

A/H5N1 has killed millions of poultry in a growing number of countries throughout Asia, Europe, and Africa. Health experts are concerned that the coexistence of human flu viruses and avian flu viruses (especially H5N1) will provide an opportunity for genetic material to be exchanged between species-specific viruses, possibly creating a new virulent influenza strain that is easily transmissible and lethal to humans.

Influenza A/H5N1 was first recorded in a small outbreak among poultry in Scotland in 1959, with numerous outbreaks subsequently in every continent. The first known transmission of A/H5N1 to a human occurred in Hong Kong in 1997, when there was an outbreak of 18 human cases resulting in 6 deaths. It was determined that all the infected people had been exposed to infected birds in poultry markets. As the disease continued to spread among poultry flocks in the territory, the decision was made to cull all 1.6 million poultry in the area and to impose strict controls on the movement and handling of poultry. This terminated the outbreak.

There is weak evidence to support limited human-to-human transmission of A/H5N1 in 139 outbreaks between 2005 and 2009 in Sumatra. The reproduction number was well below the threshold for sustained transmission.

Between 2003 and December 2024, the World Health Organization has recorded 963 cases of confirmed H5N1 influenza, leading to 465 deaths. The true fatality rate may be lower because some cases with mild symptoms may not have been identified as H5N1.

Influenza A virus subtype H7N9

Further information: Influenza A virus subtype H7N9

Live poultry market in Xining, China, 2008.

A significant outbreak of influenza A virus subtype H7N9 (A/H7N9) started in March 2013 when severe influenza affected 18 humans in China; six subsequently died. It was discovered that a low pathogenic strain of A/H7N9 was circulating among chickens, and that all the affected people had been exposed in poultry markets. Further cases among humans and poultry in mainland China continued to be identified sporadically throughout the year, followed by a peak around the festival season of Chinese New Year (January and February) in early 2014 which was attributed to the seasonal surge in poultry production. Up to December 2013, there had been 139 cases with 47 deaths.

Infections among humans and poultry continued during the next few years, again with peaks around the new year. In 2016 a virus strain emerged which was highly pathogenic to chickens. In order to contain the HPAI outbreak, the Chinese authorities in 2017 initiated a large scale vaccination campaign against avian influenza in poultry. Since then, the number of outbreaks in poultry, as well as the number of human cases, dropped significantly. In humans, symptoms and mortality for both LPAI and HPAI strains have been similar. Although no human H7N9 infections have been reported since February 2019, the virus is still circulating in poultry, particularly in laying hens. It has demonstrated antigenic drift to evade vaccines, and remains a potential threat to the poultry industry and public health.

Genetic and evolutionary analyses have shown that the A(H7) viruses in the Chinese outbreak probably transferred from domestic duck to chicken populations in China and then reassorted with poultry influenza A(H9N2) to generate the influenza A(H7N9) strain that affected humans. The genetic characteristics of A(H7N9) virus are of concern because of their pandemic potential, e.g. their potential to recognise human and avian influenza virus receptors which affects the ability to cause sustained human-to-human transmission, or the ability to replicate in the human host.

Between February 2013 and February 2019, there were 1,568 confirmed human cases and 616 deaths associated with the outbreak in China. The majority of human cases have reported contact with poultry in markets or farms. Transmission between humans remains limited with some evidence of small family clusters. However, there is no evidence of sustained human-to-human transmission of A/H7N9 influenza.

During early 2017, outbreaks of avian influenza A(H7N9) occurred in poultry in the USA. The strain in these outbreaks was of North American origin and is unrelated to the Asian lineage H7N9 which is associated with human infections in China.

Domestic animals

Several domestic species have been infected with and shown symptoms of H5N1 viral infection, including cats, dogs, ferrets, pigs, and birds.

Poultry

Attempts are made in the United States to minimize the presence of HPAI in poultry through routine surveillance of poultry flocks in commercial poultry operations. Detection of a HPAI virus may result in immediate culling of the flock. Less pathogenic viruses are controlled by vaccination.

• 
  Birds affected by highly pathogenic avian influenza (HPAI) could show swelling of the head, wattles, combs, and face. USDA photo
   
• 
  Purple discoloration of the comb could indicate highly pathogenic avian influenza (HPAI). USDA photo
   
• 
  Hemorrhaging of the skin and legs is just one of the signs birds might exhibit when infected with the highly pathogenic avian influenza (HPAI) virus. USDA photo
   
• 
  A chicken being tested for flu

Dairy cows

During April 2024, avian influenza was first detected in dairy cows in several US states and subsequently spread more widely through the year. Influenza A(H5N1) was found to be present at high levels in the mammary glands and in the milk of affected cows. It was shown that the virus can persist on milking equipment, which provides a probable transmission route for cow-to-cow and cow-to-human spread. A number of humans who had been in contact with cows tested positive for the virus, with mild symptoms. According to CDC, 7% of 115 dairy workers had evidence of recent infection in a study from Michigan and Colorado from June to August 2024 – half of them asymptomatic. This is higher than estimates from prior transmission studies in poultry. All dairy workers had worked in cleaning the milk parlor and none had used personal protective equipment.

Cats

Cats with avian influenza exhibit symptoms that can result in death. The avian influenza viruses cats may get include H5N1 or H7N2, notable pathogenic subtypes of the virus. In order to get the virus, a cat would need to be in contact with infected waterfowl, poultry, or uncooked poultry. Two of the main organs that the virus affects are the lungs and liver.

Global aspects

Global measures

In 2005, the formation of the International Partnership on Avian and Pandemic Influenza was announced in order to elevate the importance of avian flu, coordinate efforts, and improve disease reporting and surveillance in order to better respond to future pandemics. New networks of laboratories have emerged to detect and respond to avian flu, such as the Crisis Management Center for Animal Health, the Global Avian Influenza Network for Surveillance, OFFLU, and the Global Early Warning System for major animal diseases. After the 2003 outbreak, WHO member states have also recognized the need for more transparent and equitable sharing of vaccines and other benefits from these networks. Cooperative measures created in response to HPAI have served as a basis for programs related to other emerging and re-emerging infectious diseases.

Impact on national policies

HPAI control has also been used for political ends. In Indonesia, negotiations with global response networks were used to recentralize power and funding to the Ministry of Health. In Vietnam, policymakers, with the support of the Food and Agriculture Organization of the United Nations (FAO), used HPAI control to accelerate the industrialization of livestock production for export by proposing to increase the portion of large-scale commercial farms and reducing the number of poultry keepers from 8 to 2 million by 2010.

Traditional Asian practices

Backyard poultry production was viewed as "traditional Asian" agricultural practices that contrasted with modern commercial poultry production and seen as a threat to biosecurity. Backyard production appeared to hold greater risk than commercial production due to lack of biosecurity and close contact with humans, though HPAI spread in intensively raised flocks was greater due to high density rearing and genetic homogeneity. Asian culture itself was blamed as the reason why certain interventions, such as those that only looked at placed-based interventions, would fail without looking for multifaceted solutions.

Economic impact

Approximately 20% of the protein consumed in developing countries come from poultry. A report by FAO totalled economic losses caused by avian influenza in South East Asia up to 2005 around US$10 billion. This had the greatest impact on small scale commercial and backyard producers.

As poultry serves as a source of food security and liquid assets, the most vulnerable populations were poor, small scale farmers. The loss of birds due to HPAI and culling in Vietnam led to an average loss of 2.3 months of production and US$69–108 for households where many have an income of $2 a day or less. The loss of food security for vulnerable households can be seen in the stunting of children under five in Egypt. Women are another population at risk as in most regions of the world, small flocks are tended to by women. Widespread culling also resulted in the decreased enrollment of girls in school in Turkey.

F-score

From Wikipedia, the free encyclopedia

https://en.wikipedia.org/wiki/F-score

Precision and recall

In statistical analysis of binary classification and information retrieval systems, the F-score or F-measure is a measure of predictive performance. It is calculated from the precision and recall of the test, where the precision is the number of true positive results divided by the number of all samples predicted to be positive, including those not identified correctly, and the recall is the number of true positive results divided by the number of all samples that should have been identified as positive. Precision is also known as positive predictive value, and recall is also known as sensitivity in diagnostic binary classification.

The F₁ score is the harmonic mean of the precision and recall. It thus symmetrically represents both precision and recall in one metric. The more generic ${\displaystyle F_{\beta }}$ score applies additional weights, valuing one of precision or recall more than the other.

The highest possible value of an F-score is 1.0, indicating perfect precision and recall, and the lowest possible value is 0, if the precision or the recall is zero.

Etymology

The name F-measure is believed to be named after a different F function in Van Rijsbergen's book, when introduced to the Fourth Message Understanding Conference (MUC-4, 1992).

Definition

The traditional F-measure or balanced F-score (F₁ score) is the harmonic mean of precision and recall:

${\displaystyle F_1=\frac{2}{{\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{l}}^{-1}+{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}}^{-1}}=2\frac{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}\cdot \mathrm{r}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{l}}{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}+\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{l}}=\frac{2\mathrm{T}\mathrm{P}}{2\mathrm{T}\mathrm{P}+\mathrm{F}\mathrm{P}+\mathrm{F}\mathrm{N}}}$

F_β score

A more general F score, ${\displaystyle F_{\beta }}$, that uses a positive real factor ${\displaystyle \beta }$, where ${\displaystyle \beta }$ is chosen such that recall is considered ${\displaystyle \beta }$ times as important as precision, is:

${\displaystyle F_{\beta }=(1+{\beta }^2)\cdot \frac{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}\cdot \mathrm{r}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{l}}{({\beta }^2\cdot \mathrm{p}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n})+\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{l}}}$

In terms of Type I and type II errors this becomes:

${\displaystyle F_{\beta }=\frac{(1+{\beta }^2)\cdot \mathrm{t}\mathrm{r}\mathrm{u}\mathrm{e}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}}{(1+{\beta }^2)\cdot \mathrm{t}\mathrm{r}\mathrm{u}\mathrm{e}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}+{\beta }^2\cdot \mathrm{f}\mathrm{a}\mathrm{l}\mathrm{s}\mathrm{e}\mathrm{n}\mathrm{e}\mathrm{g}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}+\mathrm{f}\mathrm{a}\mathrm{l}\mathrm{s}\mathrm{e}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}}}$

Two commonly used values for ${\displaystyle \beta }$ are 2, which weighs recall higher than precision, and 0.5, which weighs recall lower than precision.

The F-measure was derived so that ${\displaystyle F_{\beta }}$ "measures the effectiveness of retrieval with respect to a user who attaches ${\displaystyle \beta }$ times as much importance to recall as precision". It is based on Van Rijsbergen's effectiveness measure

${\displaystyle E=1-{\left(\frac{\alpha }{p}+\frac{1-\alpha }{r}\right)}^{-1}}$

Their relationship is: ${\displaystyle F_{\beta }=1-E}$ where ${\displaystyle \alpha =\frac{1}{1+{\beta }^2}}$

Diagnostic testing

This is related to the field of binary classification where recall is often termed "sensitivity".

		Predicted condition
	Total population = P + N	Predicted positive	Predicted negative	Informedness, bookmaker informedness (BM) = TPR + TNR − 1	Prevalence threshold (PT) = ⁠√TPR × FPR - FPR/TPR - FPR⁠
Actual condition	Positive (P)	True positive (TP), hit	False negative (FN), miss, underestimation	True positive rate (TPR), recall, sensitivity (SEN), probability of detection, hit rate, power = ⁠TP/P⁠ = 1 − FNR	False negative rate (FNR), miss rate type II error  = ⁠FN/P⁠ = 1 − TPR
	Negative (N)	False positive (FP), false alarm, overestimation	True negative (TN), correct rejection	False positive rate (FPR), probability of false alarm, fall-out type I error  = ⁠FP/N⁠ = 1 − TNR	True negative rate (TNR), specificity (SPC), selectivity = ⁠TN/N⁠ = 1 − FPR
	Prevalence = ⁠P/P + N⁠	Positive predictive value (PPV), precision = ⁠TP/TP + FP⁠ = 1 − FDR	False omission rate (FOR) = ⁠FN/TN + FN⁠ = 1 − NPV	Positive likelihood ratio (LR+) = ⁠TPR/FPR⁠	Negative likelihood ratio (LR−) = ⁠FNR/TNR⁠
	Accuracy (ACC) = ⁠TP + TN/P + N⁠	False discovery rate (FDR) = ⁠FP/TP + FP⁠ = 1 − PPV	Negative predictive value (NPV) = ⁠TN/TN + FN⁠ = 1 − FOR	Markedness (MK), deltaP (Δp) = PPV + NPV − 1	Diagnostic odds ratio (DOR) = ⁠LR+/LR−⁠
	Balanced accuracy (BA) = ⁠TPR + TNR/2⁠	F1 score = ⁠2 PPV × TPR/PPV + TPR⁠ = ⁠2 TP/2 TP + FP + FN⁠	Fowlkes–Mallows index (FM) = √PPV × TPR	Matthews correlation coefficient (MCC) = √TPR × TNR × PPV × NPV - √FNR × FPR × FOR × FDR	Threat score (TS), critical success index (CSI), Jaccard index = ⁠TP/TP + FN + FP⁠

Type I error: A test result which wrongly indicates that a particular condition or attribute is present

Normalised harmonic mean plot where x is precision, y is recall and the vertical axis is F₁ score, in percentage points

Precision-Recall Curve: points from different thresholds are color coded, the point with optimal F-score is highlighted in red

Dependence of the F-score on class imbalance

Precision-recall curve, and thus the ${\displaystyle F_{\beta }}$ score, explicitly depends on the ratio ${\displaystyle r}$ of positive to negative test cases. This means that comparison of the F-score across different problems with differing class ratios is problematic. One way to address this issue (see e.g., Siblini et al., 2020) is to use a standard class ratio ${\displaystyle r_0}$ when making such comparisons.

Applications

The F-score is often used in the field of information retrieval for measuring search, document classification, and query classification performance. It is particularly relevant in applications which are primarily concerned with the positive class and where the positive class is rare relative to the negative class.

Earlier works focused primarily on the F₁ score, but with the proliferation of large scale search engines, performance goals changed to place more emphasis on either precision or recall and so ${\displaystyle F_{\beta }}$ is seen in wide application.

The F-score is also used in machine learning. However, the F-measures do not take true negatives into account, hence measures such as the Matthews correlation coefficient, Informedness or Cohen's kappa may be preferred to assess the performance of a binary classifier.

The F-score has been widely used in the natural language processing literature, such as in the evaluation of named entity recognition and word segmentation.

Properties

The F₁ score is the Dice coefficient of the set of retrieved items and the set of relevant items.

• The F₁-score of a classifier which always predicts the positive class converges to 1 as the probability of the positive class increases.
• The F₁-score of a classifier which always predicts the positive class is equal to 2 * proportion_of_positive_class / ( 1 + proportion_of_positive_class ), since the recall is 1, and the precision is equal to the proportion of the positive class.
• If the scoring model is uninformative (cannot distinguish between the positive and negative class) then the optimal threshold is 0 so that the positive class is always predicted.
• F₁ score is concave in the true positive rate.

Criticism

David Hand and others criticize the widespread use of the F₁ score since it gives equal importance to precision and recall. In practice, different types of mis-classifications incur different costs. In other words, the relative importance of precision and recall is an aspect of the problem.

According to Davide Chicco and Giuseppe Jurman, the F₁ score is less truthful and informative than the Matthews correlation coefficient (MCC) in binary evaluation classification.

David M W Powers has pointed out that F₁ ignores the True Negatives and thus is misleading for unbalanced classes, while kappa and correlation measures are symmetric and assess both directions of predictability - the classifier predicting the true class and the true class predicting the classifier prediction, proposing separate multiclass measures Informedness and Markedness for the two directions, noting that their geometric mean is correlation.

Another source of critique of F₁ is its lack of symmetry. It means it may change its value when dataset labeling is changed - the "positive" samples are named "negative" and vice versa. This criticism is met by the P4 metric definition, which is sometimes indicated as a symmetrical extension of F₁.

Difference from Fowlkes–Mallows index

While the F-measure is the harmonic mean of recall and precision, the Fowlkes–Mallows index is their geometric mean.

Extension to multi-class classification

The F-score is also used for evaluating classification problems with more than two classes (Multiclass classification). A common method is to average the F-score over each class, aiming at a balanced measurement of performance.

Macro F1

Macro F1 is a macro-averaged F1 score aiming at a balanced performance measurement. To calculate macro F1, two different averaging-formulas have been used: the F1 score of (arithmetic) class-wise precision and recall means or the arithmetic mean of class-wise F1 scores, where the latter exhibits more desirable properties.

Micro F1

Micro F1 is the harmonic mean of micro precision (number of correct predictions normalized by false positives) and micro recall (number of correct predictions normalized by false negatives). Since in multi-class evaluation the overall amount of false positives equals the amount of false negatives, micro F1 is equivalent to Accuracy.

the number of real positive cases in the data

A test result that correctly indicates the presence of a condition or characteristic

Type II error: A test result which wrongly indicates that a particular condition or attribute is absent

the number of real negative cases in the data

A test result that correctly indicates the absence of a condition or characteristic

Reticulate evolution

From Wikipedia, the free encyclopedia
https://en.wikipedia.org/wiki/Reticulate_evolution

Phylogenetic network depicting reticulate evolution: Lineage B results from a horizontal transfer between its two ancestors A and C (blue, dotted lines).

Reticulate evolution, or network evolution is the origination of a lineage through the partial merging of two ancestor lineages, leading to relationships better described by a phylogenetic network than a bifurcating tree. Reticulate patterns can be found in the phylogenetic reconstructions of biodiversity lineages obtained by comparing the characteristics of organisms. Reticulation processes can potentially be convergent and divergent at the same time. Reticulate evolution indicates the lack of independence between two evolutionary lineages. Reticulation affects survival, fitness and speciation rates of species. 

Reticulate evolution can happen between lineages separated only for a short time, for example through hybrid speciation in a species complex. Nevertheless, it also takes place over larger evolutionary distances, as exemplified by the presence of organelles of bacterial origin in eukaryotic cells.

Reticulation occurs at various levels: at a chromosomal level, meiotic recombination causes evolution to be reticulate; at a species level, reticulation arises through hybrid speciation and horizontal gene transfer; and at a population level, sexual recombination causes reticulation.

The adjective reticulate stems from the Latin words reticulatus, "having a net-like pattern" from reticulum, "little net."

Underlying mechanisms and processes

Since the nineteenth century, scientists from different disciplines have studied how reticulate evolution occurs. Researchers have increasingly succeeded in identifying these mechanisms and processes. It has been found to be driven by symbiosis, symbiogenesis (endosymbiosis), lateral gene transfer, hybridization and infectious heredity.

Symbiosis

Symbiosis is a close and long-term biological interaction between two different biological organisms. Often, both of the organisms involved develop new features upon the interaction with the other organism. This may lead to the development of new, distinct organisms. The alterations in genetic material upon symbiosis can occur via germline transmission or lateral transmission. Therefore, the interaction between different organisms can drive evolution of one or both organisms.

Symbiogenesis

Symbiogenesis (endosymbiosis) is a special form of symbiosis whereby an organism lives inside another, different organism. Symbiogenesis is thought to be very important in the origin and evolution of eukaryotes. Eukaryotic organelles, such as mitochondria, have been theorized to have been originated from cell-invaded bacteria living inside another cell.

Lateral gene transfer

Lateral gene transfer, or horizontal gene transfer, is the movement of genetic material between unicellular and/or multicellular organisms without a parent-offspring relationship. The horizontal transfer of genes results in new genes, which could give new functions to the recipient and thus could drive evolution.

Hybridization

In the neo-Darwinian paradigm, one of the assumed definition of a species is that of Mayr's, which defines species based upon sexual compatibility. Mayr's definition therefore suggests that individuals that can produce fertile offspring must belong to the same species. However, in hybridization, two organisms produce offspring while being distinct species. During hybridization the characteristics of these two different species are combined yielding a new organism, called a hybrid, thus driving evolution.

Infectious heredity

Infectious agents, such as viruses, can infect the cells of host organisms. Viruses infect cells of other organisms in order to enable their own reproduction. Hereto, many viruses can insert copies of their genetic material into the host genome, potentially altering the phenotype of the host cell. When these viruses insert their genetic material in the genome of germ line cells, the modified host genome will be passed onto the offspring, yielding genetically differentiated organisms. Therefore, infectious heredity plays an important role in evolution, for example in the formation of the female placenta.

Models

Reticulate evolution has played a key role in the evolution of some organisms such as bacteria and flowering plants. However, most methods for studying cladistics have been based on a model of strictly branching cladogeny, without assessing the importance of reticulate evolution. Reticulation at chromosomal, genomic and species levels fails to be modelled by a bifurcating tree.

According to Ford Doolittle, an evolutionary and molecular biologist: “Molecular phylogeneticists will have failed to find the “true tree,” not because their methods are inadequate or because they have chosen the wrong genes, but because the history of life cannot properly be represented as a tree”.

Reticulate evolution refers to evolutionary processes which cannot be successfully represented using a classical phylogenetic tree model, as it gives rise to rapid evolutionary change with horizontal crossings and mergings often preceding a pattern of vertical descent with modification. Reconstructing phylogenetic relationships under reticulate evolution requires adapted analytical methods. Reticulate evolution dynamics contradict the neo-Darwininan theory, compiled in the Modern Synthesis, by which the evolution of life occurs through natural selection and is displayed with a bifurcating or branching pattern. Frequent hybridisation between species in natural populations challenges the assumption that species have evolved from a common ancestor by simple branching, in which branches are genetically isolated. The study of reticulate evolution is said to have been largely excluded from the modern synthesis. The urgent need for new models which take reticulate evolution into account has been stressed by many evolutionary biologists, such as Nathalie Gontier who has stated "reticulate evolution today is a vernacular concept for evolutionary change induced by mechanisms and processes of symbiosis, symbiogenesis, lateral gene transfer, hybridization, or divergence with gene flow, and infectious heredity". She calls for an extended evolutionary synthesis that integrates these mechanisms and processes of evolution.

Applications

Reticulate evolution has been extensively applied to plant hybridization in agriculture and gardening. The first commercial hybrids appeared in the early 1920s. Since then, many protoplast fusion experiments have been carried out, some of which were aimed at improvement of crop species. Wild types possessing desirable agronomic traits are selected and fused in order to yield novel, improved species. The newly generated plant will be improved for traits such as better yield, greater uniformity, improved color, and disease resistance.

Examples

Reticulate evolution is regarded as a process that has shaped the histories of many organisms. There is evidence of reticulation events in flowering plants, as the variation patterns between angiosperm families strongly suggests there has been widespread hybridisation. Grant states that phylogenetic networks, instead of phylogenetic trees, arise in all major groups of higher plants. Stable speciation events due to hybridisation between angiosperm species supports the occurrence of reticulate evolution and highlights the key role of reticulation in the evolution of plants.

Genetic transfer can occur across wide taxonomic levels in microorganisms and become stably integrated into the new microbial populations, as has been observed through protein sequencing. Reticulation in bacteria usually only involves the transfer of only a few genes or parts of these. Reticulate evolution driven by lateral gene transfer has also been observed in marine life. Lateral genetic transfer of photo-response genes between planktonic bacteria and Archaea has been evidenced in some groups, showing an associated increase in environmental adaptability in organisms inhabiting photic zones.

Moreover, in the well-studied Darwin finches signs of reticulate evolution can be observed. Peter and Rosemary Grant, who carried out extensive research on the evolutionary processes of the Geospiza genus, found that hybridization occurs between some species of Darwin finches, yielding hybrid forms. This event could explain the origin of intermediate species. Jonathan Weiner commented on the observations of the Grants, suggesting the existence of reticulate evolution: "To the Grants, the whole tree of life now looks different from a year ago. The set of young twigs and shoots they study seems to be growing together in some seasons, apart in others. The same forces that created these lines are moving them toward fusion and then back toward fission."; and "The Grants are looking at a pattern that was once dismissed as insignificant in the tree of life. The pattern is known as reticulate evolution, from the Latin reticulum, diminutive for net. The finches' lines are not so much lines or branches at all. They are more like twiggy thickets, full of little networks and delicate webbings."